Everyone has probably heard of saying "history repeats," or "history repeats sooner than later," or any other derivation of it. It's true, plenty of things move cyclically in markets and repeat over and over again, but one thing to keep in mind is that they do not necessarily repeat with full consistency. Some patterns are more consistent than others, some have larger gaps in their behaviour.
Understanding the chances of how frequently the consistency fails to repeat could be an important factor in establishing a proper realistic mindset. Recency bias can often as well cloud traders' judgment where they anticipate for history to repeat exactly like it did last time, just because that last event is still hanging in their memory. In some cases that will be true and it will repeat in very similar fashion, but it is important to backtest each asset or market specifically to find beforehand if that is a realistic expectation or just an over-blown recency bias.
Think of pattern behavior and its consistency like this:
Once the pattern starts to repeat itself as it is seen on the left example with a strong consistency of "winning" behavior (on the chart, on macro analysis, or whichever else the case...) what it often creates is that it puts short term recency bias on the trader, where a trader will expect this high consistency of behavior to keep repeating without any failures in between (especially true on winning side). This is especially common to traders who have not yet traded/executed the pattern with enough trades.
While if a trader trades a pattern with performance on the right side of the image, he will not have high expectations of future performance because the pattern performs too randomly (50% reaches expectation and the rest does not). Again this applies mostly to traders who have not executed specific patterns with enough trades as with those the recency bias will play the strongest role. With experience, this short-term bias is pushed down with better realistic expectations at each specific trade since trader gets a lot wider assumption on what each pattern should do not based on past 5 examples (such as image shows) but instead the broader sample case of 50 or 100 samples, which smooths out realistic expectation better.
This creates something called in trading "short term cognitive bias or recency bias," where if the trader has had a performance on the pattern as seen on the left example, he/she might start to expect too much of this consistency to keep performing at the rates it was already showing over past few examples. While in reality, sooner or later, chances are there will be a slump period where pattern will all of sudden print a series of failures to meet expectations, which is what often happens.
The reason to point this out is that in my cases, I had had few times significant losses on trades when such strong and obvious consistency presented itself (over past several trades/patterns), and I have seen many other traders falling for the same sword, especially in first 2 years of trading. It is almost like a reality check where the market wants you to cool off to get ready for a potential slump period of bad luck, but cognitive bias in the brain is working completely against that.
Below is a good example of ticker FFHL, where the daily chart has a strong pattern of gap-spike and rejection moves. The pattern is very consistent, and yet on the 23rd of August, the pattern completely changed its behavior/performance. This is the kind of consistent pattern that usually burns many traders because they get too comfortable expecting the same behavior on an asset that has been repeated before 30 times. It is important to be flexible in trading and understand the above concept of patterns behavior distribution over 5 or 100 samples. Over 5 samples patterns might have a high consistency of behavior, but over 100 samples there will be extremes and anomalies that will go against such consistency of behavior present. The question is only at what frequency and rate, and that will depend on pattern specifically.
In my opinion, traders should avoid forming strong opinions on patterns where the sample case numbers are too low, such as only 5 consecutive patterns for example. It is common to see in trading, where beginner traders will start doubting the strategy they trade based on the 5 trades they took since they started to use it and wondered why it is not working. For any credible questioning, if the pattern can be traded with an edge, there should be a minimum of 30 samples of data (at the lowest side), but better would at least 100 samples. In trading, you are not trading hits, you are trading the process. And the process, by default, requires a minimum amount of data to be called process in the first place.
Another macro example could be Swiss franc in 2015. Consistency of previous bounces on EURCHF and the backing of the central bank by soft pegging the currency provides traders with consistent behavior on retests of peg level. But once the currency peg was removed, the move was crushing, usually, the assets that have very consistent behavior of bounces around the same price levels tend to have the biggest moves once that behavior changes.
The example on Swiss franc is not about predicting such behavior (as the majority of traders will be unable to do so) but rather pointing out how strong of a move can be, once very consistent behavior gets flipped and crushed.
The example below is my worst loss of 2018 on ticker MRCY where we had two very similar price accumulation structures, with similar behavior on tape at least initially, and I was way too biased on the second play, expecting it also to play out the same as initial first play, mostly because they were so close together and developing so similarly. Textbook short-term cognitive bias. My size was large on trade due to that, and funny enough, by the last 25 minutes while in trade, I started to observe on tape that huge buyer was trying to lift price, and he was always offloaded and stuffed big time against some seller/sellers. And that was where I started to think if this trade does not work out and this large buyer flushes out, the price might tank fast. But I did not do anything, did not size down position or cut it, due to the short-term cognitive bias, I was ready for a pattern to play out with gains on long, even that my rational thinking processes were telling me to size down. I have seen this mistake hundreds of times on other traders, yet I was also unable to react to incorrect responses. I took the loss of around 38,50 with the delayed response (entry 39,00ish). It was textbook trade of being caught with the pants down.
It is a powerful concept this short-term cognitive bias / recency bias and trying to always be humble and open-minded on setup without having too many expectations on play is always a huge help to fight this monster. Understanding how patterns behavior is distributed over time and executions helps a lot and is always open-minded on each new trade.
Distribution of successful patterns or trades
To put it in a simple case scenario. Imagine a trading setup with an 80% success rate, meaning 8 out of 10 times the setup performs the way the trader expects it, 2 times it fails to do so (over 100 samples). This would count in the trading world as a very high-edged play. And yet, even such high-edged play can have one row of consistent performance where it fails 10 times in a row. Is it possible? Absolutely. Why? Because patterns are never distributed with perfect mathematical consistent ratios like 1, 0, 1, 0,1,0. Instead they go: 1,0,1,1,1,0,1....
Thus it's always a need to be open-minded, just because you are watching the same pattern that performed exactly as you expected last time and is high performing play, it could still fail on the current case that one is trading. This is one of the hardest concepts for any trader to get used to because we all seek robustness in the business or work we do, and this concept challenges every trader every day as on every step you take one can never really be certain how it will turn out.
Below is an example of pattern distribution over 100 samples or over 5 consequent randomly collected samples of that specific pattern. This is something every trader needs to keep in mind; even great-edged patterns can over random 5 consequent taken trades to perform terribly with 5 losses in a row.
The more consistently the history repeats, the more traders will notice it. One of the most difficult things in trading is the distribution of successful patterns and managing the expectations on each play. If all traded patterns were distributed in perfect mathematical ratios (winner, loser, winner, loser...), trading would be much easier.
Or to keep it with a bit more depressing but blunt and real note: At each specific play in the market, a trader has 0 control or certainty over whether the play will work out or not. This should be used as one of the core principles within the trading, as the trade management, risk, and formulation of the thesis should all revolve around this principle at the core.