Hello and welcome to lesson eight of the advanced psychology course for high performance trading. This video is looking at the practical side of probabilities and probabilistic thinking in trading. It’s the second part of the 2 part topic in this course.
And as far as I am concerned, this is the most important video of the course. Regardless of all the other maladaptive issues we have with trading psychology, what you see here is the make or break factor. The previous video was theory based, but this one is a practical demonstration of how edges and probabilities are played out in reality, and more importantly how we perceive them playing out in reality.
I want to help you understand how system 1 functions with probabilities. Which truth be told, is not very well at all. I think this is the main reason traders fail.
# As I underlined in the last lesson, understanding probabilities in trading is not # about improving your maths or creating complex algorithms to compute and execute objective probabilities. That might help when developing your trading strategies, but day to day trading is about improving your # intuitive sense of probabilities, which is initially maladaptive for trading. The first important thing you must understand is in trading, you will never know for sure exactly what your edge is in terms of an exact percentage number.
In certain market conditions your edge may be very high, and in other conditions it’s very low. The exact same system traded during one period may give amazing results, but over another period, only mediocre or even very poor results. Neither is a true representation of the system edge.
It’s only when you’ve been trading the same system, with consistency, over many decades and through all market conditions, can you begin to get close to knowing your true edge, but you will never know for certain. But despite that, it doesn't stop us when we're trading to make continuous assessments of what our system might be in line to deliver. And it's this continuous assessment that if not monitored and understood can cause us to either mistrust our system, or put way too much trust in it.
# Because as we've learned over the last seven videos, we're born with a three-stage assessment of probability. It will, it won't, or the much more stressful anxiety state of, it might; which we try and avoid at all costs. And we try to avoid this by increasing certainty by adding more filters, tools, or fiddling with and experimenting with our process on a trade to trade basis.
So this video is all about becoming more comfortable with the maybe state by understanding probabilities. # But there's a major barrier to that which statisticians and mathematicians would call statistical inference. Let me demonstrate statistical inference for you in the simplest terms.
Using the familiarity of a coin toss which we know has # exactly a 0. 5 probability of coming up either heads or tails. And this is real data I’m using.
I really flipped these coins, and these are my first time results. It goes without saying that when you flip a coin once and it comes up heads as it did here, you don't then think that it always will be heads. # After another four flips you can see the probability beginning to play out.
# And after 10 flips you can see that we ended up with # 7 heads and 3 tails. Now after 10 flips you don't think; “hold on a minute I thought this coin was supposed to be 50 50. ” You don't suspect that this is a weighted coin.
You accept that small numbers of outcomes have huge errors or variations in results. System 1 however does make an inference. Simply meaning it infers or assumes something.
Or at least it would do if it wasn’t already so familiar with coin flips like this. # System 1's inference of probability in this case would be 0. 7; in that there seemed to be a 70% chance of the coin coming up heads.
If this wasn’t a coin, but perhaps a pebble, that’s exactly what system 1 would infer. But because you do know what the probability is with coins, and you are already familiar with this, and this is the key here; in this situation you accept that in order for that 50 50 probability to play out we need to flip the coin many, many more times, not just 10. # Okay so let's go ahead and do another 10.
# Okay # now # again # this # is all actual first time results I did manually, and here we are with a surprising string of tails. Now you might accept that, in line with your understanding of randomness and probability with coins, but let's continue. # # # # # And on now we've ended up with # two heads and eight tails.
So, system 1's inference when looking at this # run in isolation is now 0. 2. Of course, this inference is overridden, because you knew at the outset that this is 50 50 and despite what the short-term results might be telling you, the inference is wrong, and eventually if you kept on going and added all these scores up it would of course get closer and closer to 50 50.
-------live---- Let me show you that in action. # This is a great little online app from California University to bring statistical inference to life for you. It's basically an online automatic coin flipper, and speeds up the process of sitting at home for hours on end flipping a coin which I can tell you from experience is not that much fun.
# In this section we can set the probability which of course is 50 50 for coins or 0. 5. # And the number of tosses in a set, which we'll just go ahead and change # to 10.
And we'll leave repetitions of sets # on one for now. # Okay so let's go ahead and flip 10 coins. (pause) So, we've got eight heads and two tails.
Now down here on this graph we have the inference drawn from that small run of flips. So, it's put a dot on eight. Now you know it's not eight or 0.
8, the coin is 50 50. So, you accept this small run of 10 flips doesn't accurately demonstrate the probability of the coin. So, let's do another 10 flips (pause) And now we've got six heads and four tails.
Again the inference on that set is drawn as a six which is also inaccurate because of the small amount of flips. But if we keep going… and I'll just keep flipping sets of 10. It looks pretty chaotic.
Now we've got two heads and eight tails. It's drawn a two down here. And I'll keep flipping.
Now we've got seven heads and three tails. So, it's drawn a seven there. And now we've got three heads and seven tails.
And already you've got a cluster here of six seven and eight, and a two and a three. Nothing actually in the middle where it should be. Where we expect it to be.
We know it's 50 50 but we haven't even seen one instance of 50 50 yet. But if we keep going… We've got seven and three. And now we've got three and seven.
And this is my first take. I'm not fixing this. This is raw data.
Not staged this in any way. We've got two heads and eight tails. And we're not seeing a pattern, or we're seeing a very strange pattern that doesn’t match what we expect.
There's seven heads and three tails. And now we've got five. The first time we've seen 50 50 actually happen.
But if we keep going, what will happen, inevitably, is that you'll start to see a cluster around the centre. So far, we've done 140 flips. Now 15 sets of 10.
And still no pattern matching what we expect from these odds. But as we keep going you'll start to see a bell curve appear. I can speed this process up by say the number of repetitions.
That is each time I flip I do 10 sets of 10 instead of 1 set. So, let's put 10 in there, and then draw the samples. So, another go.
That's 100 flips. Another 100 flips. Another 100 flips.
Another 100. And eventually what you'll start to see is this bell curve pattern appearing. And if I put a graph above it, it shows you where the cluster is in the centre.
So, after a huge amount of flips, we can see that the probability of the coin is around the centre. It's at the 0. 5 mark.
A small sample size doesn't tell us anything whatsoever about real odds or probabilities. -----end----- The actual edge is only visible over many instances like in the graph here. In this example I’ve done over 1000 coin flips to see the true 50/50 odds.
You can imagine what system 1 infers from small clusters of trades. It will make false inferences, as the odds in the short term are so skewed out of balance. If you win 8 trades out of 10, system 1 will be making an inference that your system is producing good results and it has a good edge.
And conversely if it loses 8 trades out of 10 it makes false negative inferences about the edge in your system. It would be making a mistake in both cases. # So what have we learned so far?
# Well firstly system 1 makes incredibly fast and effortless statistical inferences, particularly if it's missing critical information or has small sample sizes. To give you an example, if you went into a restaurant just one time to find the waiter or waitress rude, you might infer that the probability of all staff being rude is incredibly high and you probably won't go back. And if you only took a couple of trades one year and netted an 80% return, you'd probably infer you had some kind of trading magic and you'd go into next year with a lot of confidence.
# Now when we know for certain that the odds are 50-50, like with a coin flip, our system 1 easily accepts that outcomes are random in the short term. It already knows what the probability is so any short term inferences must be false. System 1 is not stupid, and it does learn.
But the issue comes when we don't know what the probability is, like in trading. When you don’t know the exact odds, the inference it makes is very believable. If you knew without question the edge of your system over the past 20 years in all market conditions was 0.
7, then you wouldn't care about individual trade outcomes. You would just simply keep going, exactly like flipping the coin. But you don’t know that.
# And the smaller the sample size, the higher the error in making inferences, and this is the key point. Not only are individual trade outcomes completely irrelevant, small sample sizes of trades and the inferences that we effortlessly draw from these, have huge errors in them. And this has massive consequences, for traders who just don't get it.
Traders will look at a system's result over a very short period, and if that period is profitable or it's had a high win rate, they will infer that, hey, that's a great system. Whereas if the period that they're looking at is in a loss or it's had a low win rate, they'll infer that the system doesn't work. It’s not unusual to see traders make adjustment to their system based off the outcome of 1 week of trading, even one day, maybe even a single trade.
Let me give you an example from my own trading. I’ve traded basically the same system since 2004, and taken well over 6000 live trades in that time, so I’ve got a reasonably large sample size of trades and market conditions to draw conclusions from, and I can be quite confident about the statistics and my edge. However, the win rates from one year to the next vary greatly, as does the net profit.
I’ve had years where my win rate was as high as 62% and others as low as 35%. I’ve had a year where my return was only 20% and another where I made over 130%. If I zoom in on a small scale, I’ve had months where my win rate was 10%.
Other months where it was 90%. On a weekly level, I’ve had 100% win rate or 100% loss rate. It's the act same system, but if an amateur trader looked at any period in isolation, they would infer, completely falsely, that they where different systems, simply because they don’t understand probabilities or random distribution.
The smaller the sample size of data, the less reliable the data is, as we’ve just seen with something as simple as coin flips. Probabilities and outcomes are irrelevant and meaningless in the short term. But think about what this means for traders.
I’m profitable, and have been every year since 2004, because I keep flipping the coin again and again and let my edge balance out. I ignore the short term variation. But most traders see a series of say 10 trades, and conclude it is or isn’t working.
They change something that “would” have made them the most money over those 10 trades. And then go again. But then the next 10 trades give a totally different result, so they change things again.
Then the next 10 are different again, so more changes are made. What they are essentially doing is, starting from scratch, every few trades, with essentially a different system. And they’re never consistent with any, long enough for the odds and edge to balance out.
This lack of understanding of probabilities, dooms them to failure forever. # So, let me welcome to my probability lab. This is an experiment that I set up to demonstrate what happens when we move away from the 50-50 and into unknown probabilities like trading.
I created a coin with an edge. # Coin A is a regular 50-50 coin and outcomes are completely random. But coin B has been bent slightly to give it a slight bias for landing on heads.
Although I can’t say exactly how likely. The exact edge is unknown, but there is certainly an edge. # But in bending the coin and creating that edge, I've created some expectation for a particular result.
I now expect to see more heads if I flip the coin multiple times. Now, this expectation turns out to be very important and makes our false statistical inferences even worse. Let me show you with a simple experiment.
# I flipped coin B, with its edge, in sets of 10, 10 times. So, 100 flips overall. And again, these are actual first time results.
There's no staging here. What happened, happened first time around. # So, on the first set of 10, the inference was 0.
3, which is odd because I’m expecting more heads, but actually more tails have appeared. So, if I was betting $100 each way, I'd be down $400. I've won $300 and lost $700.
# Another 10 flips, and I’ve got 50-50 and still no edge, no advantage. It already seems unfair. This is the equivalent of taking 20 trades.
Surely the edge would have been visible by now, but I'm still down $400. # On the third set of 10 it’s 0. 3.
So, we flipped it 30 times so far, and it would appear that tails is more likely. At this point, I genuinely wondered if I'd bent the right way. I was honestly confused and double checked.
And to make matters worse, I'm now down $800 had I been betting on this. Now imagine if that was trading and you purchased this system from a seller, you're 30 trades in and you've lost 19. It's totally understandable that you'd be freaking out by now.
Where is the edge you were promised? But this is the difference between what system 1 expects, and reality. Statistically speaking, 30 is just not enough events to see the edge.
And that's the problem. The inference that we make out of these small numbers is false, and it would send us into all kinds of evaluative chaos, and the $800 loss will fuel that potential chaos. And remember the time factor.
For an intraday trader, 30 trades may only represent 1 week. For a longer term trader this could be 6 months, even a year. Be honest with yourself as to whether you would stick with this and keep going without any changes based on these results.
Would you really trust this has an edge if it was trading? But we need to carry on flipping in the same consistent manner. # With another 10, at last.
That’s more like what we expected, 0. 7. I've made back some of my losses and now I'm only down $400 in total.
# And another set brings us another 0. 7, which nets me another $400, which breaks me even. # And now a 0.
8, which puts me in profit for the first time ever with a $600 profit overall. That's 60 flips, or 60 trades with a guaranteed edge and trading consistently, so to speak, before I even saw a profit. Try and imagine if this was you taking trades.
So, those last three sets of 10 met our expectations. So, is the coin biased by 0. 7 or 0.
8? At this point, recalling the last few set of flips, we might even start feeling confident that actually the edge is now here to stay and it's a big edge. # But with another 10, we made nothing.
# And then it’s back to a losing set, losing us $200 with an overall profit down to $400. # And another break even, which now seems like part of a losing streak. # And in the last 10, we've got a very slight edge of heads, leaving us up on this set in total by $600.
Now, the reason I'm showing you this is to both demonstrate how edges play out, but also to show you what an emotional rollercoaster System 1's statistical inferences put us through. Especially when we expect there to be an edge. It's that expectation to want to see the edge playing out in the short term that causes the anxiety and the confusion when we trade.
The only real way of knowing what the edge of this coin might be is to aggregate those scores. So, let's do that. With 100 flips, that coin would seem to have an # edge of 0.
53, which is slight, actually lower than I expected. But is that it? Does this coin have a 0.
53 edge? Well, to find out, we need to keep going. The good news is that I did this for you and flipped the coin another 900 times.
Yes, it was painful, thanks for asking. What I saw was that the edge settled down more and more. And I ended up feeling pretty confident that the edge of this coin is actually 0.
65. And on that basis, I would be happy that if I could bet $100 each time and play the game infinite number of times, I would be infinitely profitable. ------live---- And if you want to do this kind of experiment for yourself, which I highly recommend that you do because it will help teach your system 1, you can use this online app for that as well.
Last time, here we had the probability set to 50-50 in single sets of 10. And when we do that, you can see # that it represents the outcomes with coins. But if we change the probability to # say 0.
65, which would be a coin with an edge, or a trading system with an edge, then we can simulate how we might react to that expectation of an edge. Okay, so let's go ahead and simulate 10 flips or trades with an edge. # You can see now that the visual representation is different.
It uses a clock hand to create the randomness and colour coded portions to represent the edge. So, the 0. 65 edge is represented in blue.
And it's more likely to land on blue, which it calls successes; or a win as we would know it. So, now we have expectation, and we know that there is an edge, and we want to see evidence of it. But the first set of 10 here didn't show it.
And you can see the inference point on the graph below says it's 0. 4. So the exact opposite of what we expect.
Oh dear, if that was a trading system, a new trader would very likely give up there and then and probably voice his frustration on a review site or forum. But anyway, let's try another 10. # Okay, so now we've got eight wins and two losses.
Even with a definite edge of 0. 65, each set of 10 flips, or 20 trades, does not represent that edge, which is why reacting to a set of outcomes is pointless, let alone reacting to an individual outcome. As I keep saying, but hopefully am now demonstrating, individual trade outcomes are meaningless.
Even small sets of trade outcomes are meaningless. But let's go ahead and ramp it up and do more trading. # Again, just like the 50-50 demonstration, you can see inference points appearing on this graph.
But slowly, after many transactions, you can start to see a pattern appearing. # And I'm going to accelerate it even more by again putting this up to 10 and then flipping 100 at a time. So here we go, # another 100, # another 100.
What's the inference? Can we start seeing what the probability is? Well there we go, we can start seeing the shape appearing.
And if I put the graph over the top, you can see the data is saying that the probability is actually playing out about the 0. 65 level. And that's after 670 flips.
That’s a lot of trades before you begin to see your edge. And 65% is a big edge. Hopefully you can see what I'm trying to demonstrate here.
Single outcomes, small sets of outcomes are meaningless. In order to see what the edge really is in reality, and see it in your equity curve, there has to be large sets of information. # But now let's get back to our two coins.
With a regular coin flip, you know the chances are 50-50. And you accept that the outcome is random on a flip by flip basis. # And you accept that expecting a particular outcome on any flip is totally ridiculous.
Your intuitive sense of probability gets this. It’s familiar, # But with coin B, we have created an edge. And remember, an edge creates expectation, which stops us accepting that the outcomes are still in fact random.
There is now a system 1 expectation that we should be able to see it with our own eyes very quickly. As traders, that means seeing it this week, or in this month's equity curve. Now, yes, intellectually we claim to accept losses and periods of draw down.
But there's always something niggling away in the back of our mind. Inexperienced traders, expecting to see an edge, are expecting it over a very small sample size. And when that doesn't happen in the short term, that's when the doubt sets in, and it sets in pretty quickly.
# To summarise so far. # If the exact odds or the edge are unknown, like in a bent coin or in a trading system, but we know that they are in our favour, we create expectation that individual or small sets of outcomes will demonstrate those favourable odds, but they don't. # Which means, despite having an edge, we are confused when the outcomes are negative and this causes pain, distrust, and total chaos in the way that we evaluate our trades as we try to work out what is going on.
# Well, now let me introduce you to a new coin. We've already discussed coin A, which is 50-50 and comes with no expectation, and coin B, which has a slight edge and triggers some unhelpful expectations of the next few outcomes, even though outcomes are still random. So, now meet coin C.
# As you can see, this is totally bent over, giving it an overwhelming edge. We don't know what the edge is precisely, but it's a big one. This is the type of edge or trading system you wish you had, or certainly the one system 1 wishes you had.
In terms of trading, this is what most system sellers are promoting. But the funny thing is with this kind of edge, is it may give you a high win rate, but overall these are losing systems. It’s not so much a high edge, more a high win rate to disguise the total lack of a real edge.
You can achieve this by having a wide stop loss and a small take profit. That will instantly give you a high win rate. But there's a very good reason that people chase these kind of systems, because, there's no pain associated with false statistical inferences, and system 1 expectations not being met.
These systems give system 1 what it wants in the short term. # Let me show you, with the results of me flipping this manually in sets of 10. # Well, the first set meets our expectations.
There it is. The edge we expected, straight out of the box and I’m $400 up on the first set. # Second set, slightly lower.
# And another. But it’s still an edge. We'd be $800 up by now.
Great. Next we’ve got # a bumper 0. 9.
And it continues, # 0. 7, # 0. 9, # break even at 0.
5, and # another # three # profitable sets. So over 100 flips, the coin seems to have an # edge of 0. 72, giving us a profit of $4,400.
And most importantly, no inference pain along the way # because our expectations have been met. Yes, we accept some losses, but system 1 was happy all the way because the edge was played out in much smaller sets. This is the coin that represents the edge all traders wish they had, or in other words, what system 1 would really accept with comfort.
----live---- # Now, if we nip over to the app and just put in coin C, I change this to 0. 72 and take 10 trades or flips, you can see again the edge is visible on the first set. Here we've got 10 wins and no losses.
Let's try again. And yes, it falls in line with expectation. Seven wins and three losses.
And again, nine wins and one loss. Eight wins and two losses. So far, no pain.
System one accepts this entirely. There’s seemingly not a big difference between 0. 65 and 0.
72 on paper, but every set of 10 here feels painless. We feel like we're on top of it. You can see what I'm trying to demonstrate here is that the lower the edge, the closer to 50-50, the bigger the pain in inferring outcome, and therefore the bigger chaos when trying to analyse short-term and individual or small sets of results.
---end---- # Okay, so where are we? # Well, of course, traders do on the face of it, accept that a trading edge is probabilistic and that you must accept losses. But system 1's expectation is that a trading edge should behave more like coin C, which is almost painless on a trade-to-trade basis.
, and certainly when we recall sets or groups of trades, because it sees the edge, very quickly. But the reality is what we experience and should ideally be able to cope with, is a lower, more realistic edge that plays out only over a much longer period. And that is what your trading system is.
It's a small edge that requires consistency of execution over the long term, despite the uncomfortableness of the statistical inference that system 1 seems to make on a day-to-day, or even month-to-month basis, and the emotional roller coaster that comes with it that I described earlier. # To drive this point home, if I gave you a guaranteed secret system to switch the odds very slightly in your favour on a gambling machine, would you expect a particular outcome on any particular pull? No, of course not.
You would keep playing knowing in the long run that you would be up. You’d know not to expect anything specific on individual goes, but you’d be confident in the edge over the long term. This is exactly how casinos work.
They don't get hung up on individual outcomes, and never panic when they lose to a player. They only care about spinning the wheel, rolling the dice, or dealing the cards over and over for as long as they can. That’s how the house advantage plays out.
If you trusted your trading edge like a casino does, then trade outcomes become inconsequential. Assuming of course you have a system with a real edge in the first place, and you have the mental capacity to trade that system consistently. # So, the reality is that indicators, tools, systems, rules, inevitably cause expectation.
They are often seen as outcome predictors rather than probabilistic advisors. And if you fall into that trap, # thinking that any part of your trading system is an outcome predictor, then you will have some expectation as to what the outcome might be. And as we've shown with the coin experiments, expectation is only going to get you into trouble, because if the expectation # is not met, the trade does the opposite of what your system supposedly predicted it would do, # then all you have is pain, disappointment, frustration, and the beginnings of system distrust.
And even if your # expectation was met, the trade did do what your system supposedly predicted it would, then you have something even worse, which is # false correlation and learning. Meaning you start to think that if you did the same again, you would get the same result. False hope.
And you might even think that result was down to your trading skill. # Your system is not a trade outcome predictor. If you see it as a predictor of individual trade outcomes, you set up expectations about the outcome of the next trade, and that will cause you nothing but pain, frustration and disappointment.
Thinking of casinos again. We say the house never loses, but they do in the short term. They lose close to 50% of every game they play.
They only have a very small edge. But they don’t care, because they have no expectation or prediction for single spins of the wheel. They know it would be stupid in the extreme to try and predict the outcome of the next spin.
They simply know they have an edge, and they know it will work if they give it enough time and opportunity, so they stay open 24 hours a day, 365 days a year so the edge can balance out and in the long term they will make millions. That’s the approach all successful traders need to have. # Okay, to summarise the last couple of slides, trading with individual or short term expectations will lead to trouble.
# If you have any expectation on trade outcomes, and they are met, then you get false correlation, confidence, and worthless analysis. This is not feeding your intuition good information. # And if your expectations are not met, then inevitably you get pain, frustration, growing distrust, paranoia, and all kinds of other things.
# The long and short of all this is, there are six things that you must accept if you are to continue to trade with real risk capital. # Firstly, your trading edge, may be completely indistinguishable from 50 50 on a trade by trade basis. It will likely not be obvious at all.
# Secondly, you must accept that no part of your trading system, no tool or indicator, predicts direction or outcome of the next trade. And to see it as such creates expectation, which, as I've said, will lead to trouble. # Thirdly, the direction and outcome of all individual trades are completely random, in the same way that the outcome of even the weighted coin B flips were random.
Of course, there will be a positive expectancy over many trades, but nonetheless, individual outcomes are random # And therefore, inferences made on individual and small sets of outcomes are all false. You will make inferences because it's an automatic, effortless system 1 process, that you can’t switch off, but you need awareness to recognise the inference is false. # And there is no direct cause and effect between your system and the outcome of the trade.
# And lastly, you need to remember that system 1, and indeed the majority of the unprofitable trading community for that matter, will initially tell you that everything I’ve said is wrong, which is why so many traders never accept the true nature of randomness, low edges and probability Okay, with all that said, there's one last thing that we need to talk about. As I mentioned earlier, there are systems out there that have very high edges, or more accurately, very high win rates, but which ultimately are losing systems that wipe out your account in the long run. System 1 craves certainty.
# So, welcome to coin D. This coin might represent a couple of things; Either a system with a crazy high win rate because it uses averaging in, no stop loss, martingale betting strategies or grid strategies. So, a system that wins nearly every single trade, but… which, when the unlikely loss happens, it’s so big as to wipe out the account.
Or it could represent a black swan event. Something so rare as to be assumed to be impossible. When we get used to success, we start to assume that something with a very high or low probability, depending on how you look at it, is actually certainty.
We get complacent or overconfident. In fact, when I made this coin, I was convinced myself that this was actually a 100% certainty coin. But I tested it nonetheless to see how long it would take to land on tails, if at all, and importantly, see how I would react to so many flips with this kind of coin.
And this is what happened. # We join the video at flip 622. Now, after 500 flips, I remember losing all expectation that it would ever land on tails.
Every single toss landed on heads. But I decided to keep going to 1000 and then give up and then make a slightly less biased coin. But anyway, incredibly, after 632 flips, it did land on tails.
If I were actually betting on this coin, after a few hundred flips, I would have been completely confident to put more and more money on it. After 500 flips, I would have bet almost anything I was so sure tails would never happen. The intuitive inference made by system 1 on these kind of odds is one of complete certainty, which breeds recklessness, or leads us to think that we don't have to plan for the improbable.
# It's like playing a giant game of Russian roulette with a thousand barrels. And this is exactly what causes catastrophic failure and losses, not just for amateurs but also for the supposedly smartest people on the planet. I’ve told you before about long term capital management.
They had 2 Nobel prize winning economists and the former vice chair for the federal reserve on their team. And they created an algorithm that in testing, had a 99% win rate. Which they inferred was the same as 100%.
And 55 major banks and hedge funds agreed with them, and they ended up with over $100 billion dollars invested with them. And they were so certain, they leveraged themselves up to the point they were actually managing over $1 trillion dollars worth of investment. They were certain, with a 99% win rate, they couldn’t lose.
That nothing could go wrong. And guess what? That unlikely event they thought couldn’t happen, happened.
They lost literally everything and nearly took out the entire global economy with them. They are the perfect demonstration to prove everything I’ve said in this video. To start with, when designing their system and coding their algorithm, they only back tested 4 years of historical data.
Which is way too small of a sample size. It’s incredible to me they thought that was sensible. And even with that, and their hypothetical results of 99% accuracy, they misinterpreted 99% for 100% certainty.
They didn’t understand probabilities at all, Nobel price wining economists; and it blew up in their faces. Even the famous George Soros, one of the greatest traders ever, invested with them and lost $2 billion in the fiasco, so don’t be too disheartened when you make similar probabilistic mistakes. We’re all vulnerable to this because of how system 1 has evolved and works.
Interestingly, Myron Scholes, the Nobel winning economist responsible for that catastrophe, went on to set up a second hedge fun after blowing up long term capital management, and lost another $600 million in the 2008 crash, so he didn’t even learn his lesson. We humans are a stubborn bunch. As a retail trader, you don’t have the luxury of playing fast and lose with other rich peoples money with no consequence if you lose it all.
For you, you have to learn this stuff if you want to succeed with your own account, because it’s your money. Win rate and high probabilities can give you the illusion of certainty. It can make you complacent.
It can make you ignore the risks and the downsides. Always remember, if something can happen, it will happen. # So, anyway, there we have it.
Here are all the coins lined up. We have coin A, which is # 50-50 and doesn't create any outcome expectation. So, with coin A, system 1's assessment of probability accepts randomness.
# And then we had coin B, which is most likely to represent a typical trading edge within a good system, which, because we know it has an edge, does create some kind of outcome expectation. And of course, when we have outcome expectation, we get all the false statistical inferences and find ourselves on an emotional roller coaster. And actually, at an individual outcome level or small clusters of outcome level, # these two coins feel very similar.
The edge is not visible… not to system 1, at least. Now, coin C, this is what system 1 really wants, but it's unrealistic. Our intuitive sense of probability behaviour is perfectly comfortable with this coin because we can see the edge in real time.
We accept some losses, but the wins are clearly more evident, so we don't freak out. But as I said, in most cases, this is a fantasy edge. It's not realistic.
This is more or less what the community refer to as the Holy Grail. # And of course, coin D, which creates false certainty and causes us to not plan for the worst, by maybe not protecting our account or piling on more and more risk. When we step back and look at this, we can see why traders have such difficulties.
We mistake probabilistic systems, for being predictive systems. We think that having the odds in our favour, somehow predicts the individual outcome, and when that doesn't happen, we think something's wrong. We mistake coin B for coin A because in the short term they are indistinguishable, and therefore attempt to create coin C, but overshoot and create coin D because of our deep desire for certainty.
Coin D in terms of trading systems are systems that use averaging down, martingale, grid strategies, wide or no stop loss, and quick profit targets. They have exceptionally high win rates, but they mistake high probability, for certainty. And in fact, these systems are always negative expectancy systems.
They always lose in the long term, and when they do, the loss is catastrophic, and the account is blown and the trader loses everything. The frustrating thing is, Coin B, or trading systems with that kind of edge, are super easy to find, and it will make you rich. But it doesn't look that much different from coin A, and as such, most traders tend to throw them away as junk.
You’ve probably had several systems like coin B, and you’ve given up on them. # Just to be clear, the point of all this is not so that you can work out what your edge is exactly. That is unknowable, and of course it will fluctuate with market conditions.
But if you have a reputable system with a positive expectancy, over time, that edge will play out as long as you stick to it and keep playing the game. And as your trading intuition grows, that edge will be extended. What I’ve tried to do is demonstrate why reacting to statistical inferences in the short term, will lead to over or under confidence in your system.
You can't turn these inferences off, but you can learn to not react to them. You can learn to see them as maladaptive advice from system 1. So, like always, it’s all about self-awareness.
# Look for and become aware of where system 1 might be struggling with probability. To what extent are you or are you not accepting that trade outcomes and small sets of trade outcomes are irrelevant? Start noticing whether your emotion is driven by whether the trade won or lost, whether you made or lost money.
In other words, are you focused on short term outcomes. # What you want is oversight to be able to discount the false inferences, and begin to place your focus on the process you are following. Trading is all about measuring your process, not outcomes.
You can’t control the outcomes, but you can control what you do. As usual, here's a quick summary slide if you want to pause the video to take notes. But where are we going with all this?
Based on everything that we've talked about today and in the last lesson, the obvious question arises about how you can assess a trading system's edge. How do you know if you've got a coin B and should stick with it or if it's just coin A and should actually be abandoned? Well, that's what I cover in the next video.
The next video is about trading systems. The next 3 videos are actually all tied closely to this topic of probabilities, in context to how we can apply it practically to our trading. I’ll look at assessing profitable strategies with genuine edges and positive expectancy, and the processes of executing those systems effectively and improving your performance.
The final phase of this course really looks at your trading performance, process evaluation, and developing feedback loops that will lead to incremental improvements in everything that you do in the markets. Make sure you click through to watch the next video as soon as you can, and if you’ve got questions, please use the comments section and I will try to reply directly, but also use your questions to create follow up videos after this course to help you make sense of the information you’ve got so far. So, thank you, and I’ll see you in the next lesson.
