Just to be clear - this is live equity for this system. This is not hypothetical return. My logic is - nothing that makes 38% return in 6 months with 4% Max DD is sustainable. Just my personal common sense. That's like 9.5 return to risk ratio... Maybe Renaissance has that kind of RR, but I'm no Renaissance. Seems like we're just seeing extremely favorable conditions for this strategy, and they already lasted longer than I would expect. System is ripe for a good DD. Val
I was curious as to whether you had done any kind of backtesting before you took this system live. And, based on those test results, what you considered realistic expectations for this system. Both in terms of returns, as well as for DD.
SPY from 4/1-9/1 had 42% gain and 7% max DD, so extremely favorable conditions for any long strategy. Yours did well to capture most of that with smaller DD and in particular to avoid the current SPY DD since 9/1.
I’m glad it’s so easy to pull this as RealTest lets me edit list of columns as a text file where each line is "name: formula” + it allows me to run a test breaking it down by a selected time period, so stats can be analyzed in more details. Here is backtest’s RR for this strategy since 2005 broken down by year with RR included. A quick note, for the RR formula I used Reward/Risk and not Risk/Reward, so it is easier to read. There are many ways to read this table, based on that data + much more going back to 1950 + some discretion here is how I interpret it: 1/1 RR (Reward/Risk) is likely over 10-20 years 2/1 RR is not unlikely over 5-10 years 4/1-2/1 RR can casually happen in any single year Having 10/1 Reward to Risk over last 6 months is pretty unusual and is certainly is at the top range of even theoretical best performance for this strategy. Absolute % return is pretty crazy too considering it used ~28% of capital on average intraday, and only had average 12% overnight exposure. Val
Wanted to do this at the end of month, but time is a sparse commodity and I’ve got some now. Pros know that neither $ or % profits mean much. There are just numbers to impress gullible or uneducated on this subject. They are the shiny object that distracts. One of the things that actually matters is Risk to Return. While it is debatable to what is a good measure of such, I’ll use Return to DD here to start with something. Btw, it is “debatable" because there could be “not materialized” risk. I’m not sure what’s the fancy way of calling this, but one example would be selling options and cashing premiums for a year without much of DD and then blowing up in one day. With stocks, over the large number of trades without averaging down and using fixed positions sizes - big part of risk will eventually materialize in DD. That’s why I thought is reasonable to start with it. As mentioned in my previous post, I do believe that the longer one trades the less that Return to DD ratio is gonna be. So this year’s result doesn’t mean I have some sort of holy grail, but it looks damn good to me and I will take a moment to enjoy it In fact it is so good that I myself, would be very skeptical of another individual trader achieving it without substantial proof. But again, I don’t think it will be this way over let’s say 5-10 years. More possible reasons why my results should be somewhat discounted: This is year has exceptional volatility There is always some element of luck when results are exceptional Important note: my reconstructed equity from trades doesn’t include certain commissions such as borrowing rates, which start being visible on a 6+ months periods. So when I do analysis of trades/equity/DD - % numbers might be off by around 10% of their value. For most practical purposes - it is not a big deal. But it will create some discrepancy between my end of month PL (based on broker statements) and long term graphs generated by my software. Net profit YTD: +93.4% (+83% calculated by IB) Max DD: -13.02% Return/DD: 7.17 Return(year adjusted)/DD: 11.09 Sortino: 4.08 # of trades: 1204 Average capital use (overnight): 37.32% This is for combined equity=whole account, which consist of results of multiple systems. Some work better is some conditions, others in other. Achieving those kind of high Return to Risk number is exactly why one should use multiple non-correlated systems. That doesn’t guarantee but statistically increases our changes to have higher return to risk. Comparison with SPY looks just ridiculous this year but I’ll indulge and include it To give you an idea on density of trades here are open positions daily: The next one is probably gonna be most controversial. As it is what most people would consider a good example of a system that should not be traded. Because visually, there are much more negative outliers than positive ones. But ultimately it doesn’t matter. Density of PL matters. A human eye can’t size it up correctly if any statistically valid sample size is used. Talking about this might take another long post, for now I’ll just use it to remind that common wisdom is very contextual and doesn’t come with a manual on when to apply it / limitations. Ultimately - beyond density, understanding implications of such PL distribution matters the most. MonteCarlo outcomes on fixed position size of 7%. My actual size varies by strategy but this is somewhat useful. The whole MonteCarlo thing is not something I normally look at. Maybe once in 6 months. There are stronger rules that govern my risk management and I believe they are more practical. Again, there could be a separate post on that. But it is always nice to see MonteCarlo simulation results looking reasonable on a large number or trades. Val
can you post your real-time entry and exit signals? I often has doubt on mechanical systems using combination of traditional indicators. your results so far are impressive, which makes me rethink if my biased view is correct.
They will be all over the place. Some are super precise, some are really bad. I don't even look at them. And when I did - I would often want to override the system because something doesn't look "timed" correctly or "out of sync with the market". Mechanical trading is about robust model and perfect or near perfect execution. Individual trades mean very little and are a distraction. Which very counter intuitive for many. I had a friend looking at my spreadsheet with individual trades PnL, after staring it for 30 seconds he said - I don't see how that would make you money, did you make any? As results are often all over the place with occasional streak of unusually high number of winners or losers. Fun fact - talking about the best ones - on multiple occasions, I've had my executions outside of official exchange OHLC EOD data. So officially, they never happened. If I post them here, very few people will even believe that is possible I'm attaching an example of a big winner and a loser. Hope it helps. Val