Say a system has win probability of 60% with 1:1 On an average let’s say this system does 5 trades a day. On one day let’s say the trader gets all winner 100% win for that day. Then next day what is the probability for win is it still 60% becuase yesterday results have no bearing on todays result or worse because law of average needs to kick in to achive 60% win as a whole. On swing trading trades seems like continuous set so I never had this question but in day trading everything kinda resets every day.... so should day trading each day trades should be treated as different sets or one continuous set
There is no correlation between each trade, every single one has 60% probability. If you flip a coin you always have 50% chance for heads and 50% for tails. If you flip twice you have 50/50 in every flip, yet only 25% chance for getting two heads in a row.
although 60% is 60% and yes should apply over all if the 60% is correct for every day on average, some days will ofcourse fare better and worse, bringing you to an average of 60%. Also, markets kinda trade the same for periods of time, so actually the conditions yesterday if great might of been 80% and the odds are that'll bleed into the same for that day and days maybe even weeks and months after. Then you'll get periods of 40% profitable cause it's choppy and dead or trendy depending on which way your trading. But over all the average if 60% is correct, will come back to 60%
I would go even further. The coin is not your PnL. Compare the two probability density functions. They won't be the same. Trading is not replicating the underlying. Otherwise we would simply buy and hold. A coin has a normal density function. Financial markets are log normal density function. Serial correlation is an indication of path dependence. Meaning that there is kind of a correlation through time. A winner would be more likely after a winner. A loss after a loss. A coin is a bad example. For we know it's 50/50 until proven otherwise. An investment strategy has no logical prior probability. The only logical prior would come from the risk reward ratio. 1Risk 1Reward: 50/50, 2Risk 1Reward:33/66 ... IMHO you need to update, take into account, day by day. Only if it's the same strategy, plan, rules. Update it.
The most simple way is check the correlation with the previous trade, or 10 trades or day or week, etc. As a simplistic example. If your strategy does better in trending market and the market is more range bound/ dead on a day, you would expect a lower than 60% win rate. So if you have profited on 1 or 0 of your first 3 trades, it could be a good signal that the rest of the trades for the day have a negative expectancy.
I test in different way, what is losing percentages, and drawdowns, because losing is only side I can somewhat control. Every trade is 50/50, and then the next one is 50/50 etc. Buy the percentages change usually by the week and symbol because of reports due, earnings, currencies, output like in Crude, so volatility up/down changes, you might have safeguards where too much or too little you don't trade, all are taken into consideration of weekly losing percentages. Also, keeping hourly percentages, you might find over a few years of data during certain times to not trade cause losing % increases or R:R too small.
Have you all read about rescaled range (R/S) analysis, with values between 0 and 1. A value of 0.5 means a random walk. A value greater than 0.5 means some nonrandom movement is present. A value less than 0.5 means movement is more reverting than even a random process would suggest. Daily movements of most financial instruments are on the order of 0.55. This means they are mostly random, but not perfectly. Tip: try looking at monthly prices. So yes, serial correlation does come and go in assets prices, but probably not predictably enough for you to develop an edge or positive expectancy trading system around. Especially at the retail level.
%% Exactly mm; even more so in the 1st quarter+ last quarter, for stocks. BUT what he may or may not run into;a 5 minute chart has much more noise, than a profitable monthly candle chart. Its not that its random, even with a coin flip ;a silver coin or 200 year old copper coin are not random prices @ all. LOL Of course with a good trend day like up on JAN 28 or like down JAN 31 , even a noise chart like 5 minutes could make money. [And as practical matter ,a 100% win rate would be more common for a market maker or investor.And even a HI% trader like Mark Weinstein did not do 100% for long, except in his trade win contest win for 3 months .Amazing]...................................................................................................................................