%% TRUE, I have what thy call ''flag patterns'' in my trading books; but i much prefer something more objective like a 200dma........ NOT saying I'm against US or UK flags LOL Its like WO'N [IBD founder] noted, ''new traders want to much, with too little work'' NON Wall Street business can commonly average profits every month; But seasonals can apply anywhere . Profits every week is not very good, with a lack of a spending plan. But my favorite contract business may have had some slow 4th quarters; no problem with a good budget.
Thanks very much for this. I think you make some good points. At the moment I am trading discretionarily to see what works and what doesn't. Until recently I thought I could just try systems with different parameters until I found the holy grail, and of course that is fraught with problems. A hybrid approach sounds plausible to me; using data and recommendations from a system, but then trading it by hand in the moment. Certainly something worth looking into.
I may make an expanded post here at some point, just dipping my toe in this community. I have worked with a couple of math PhD's and a couple of sharp developers to do exactly this. We started in crypto in 2015 part time, went full time in 2018 (BAD timing for crypto), pivoted to traditional markets in 2020 (banks and funds would not touch crypto, no surprise in hindsight), settled on futures trading in late 2021, and went live in May of 2022. The following results are 100% LIVE automated algorithmic trading (our most conservative portfolio) through Interactive Brokers: The following are several accounts we've started trading the above system in this year (some set for more aggressive trading than others). Screenshots are Interactive Brokers "Realized Summary" statements: We only trade ES contracts. 100% automated, I NEVER place a manual trade or override an automated trade signal. So, it's possible.
Amazing results. Out of curiosity, what is the basis of your algo? Is it just crunching numbers on prices and volume? Any type of cross market referencing or looking at options to help gauge market direction?
What was the peak-to-valley max. drawdown here from May '22 on ? And how many points or ticks is your average trade ?
Hi. We use AI models (evolutionary computation, genetic algorithms and several other "classic" machine learning models, not the current LLM craze) that have access to over 300 technical indicators to discover algos to try. Obviously there is EXTENSIVE and pessimistic vetting/backtesting, and out of the millions a month that can be generated, one or two might graduate to the next stage of running forward in paper trading. An algo goes through a LOT before it ever sees one real dollar in trade authority. But to answer your question, at this point, our algos are strictly TA based. We do look cross-market (sectors, large ETFs) for correlations, but those are still measured using TA only. We have looked at sentiment as an indicator (we used to half-jokingly talk about creating the TTI "Trump Tweet Indicator" ), as well as put/call spreads, but we have not seen improvement for our style of trading (most positions are held less than six hours, with dynamic but tight stop loss and take profit targets). Our algos are also tuned every weekend on multiple timeframes going back as far as Sept 2021 currently, with more recent timeframes more heavily weighted.
Our Max P2V DD was 14.9% in Sept 2022. We DO (though not since Feb currently) see periodic drawdowns approaching the 10% range, probably ten times or so since May 2022. Our average trade (we only trade ES E-Mini futures) takes profit after $7-$12 of favorable price movement. As a side note, we have tried using trailing stops for our exit, but over time, our strats seem to always do better taking profit in a fixed (and pretty tight) range rather than "letting it ride."
Do you mean $12 per 1 contract ? So less than 1 tick as it is $12.50 on 1 contract ES ? Did you factor in slippage and comms too ? Second what time frame you use for backtesting ? 1 minute or tickdata like second bars already ?
Sorry, I mean $7-$12 of underlying price movement of the price of the current quarterly ES futures contract (currently $4526), or 28-48 ticks. Our average profit (ES = $50 per $1 (or four ticks) of price movement) per trade is $350-$600 per contract traded. Our modeling and backtesting do factor in both fees/commissions and slippage. The nice thing about ES volume is the OB spread is is pretty much ALWAYS one tick (unless you're placing a $20M+ market order), so that's what we backtest with. We also factor in not assuming a trade fills in backtesting or paper trading unless the bid/ask price has EXCEEDED our target price by one tick, to avoid optimistically assuming an order is filled just because our target price is touched for a split second. Factoring in fees/comms/slippage and pessimistic fills (bid/ask exceeds target price) during strat discovery and validation weeds out a lot of higher frequency algos that look good until you factor those real life variables in. We learned that lesson the hard way a couple of years ago. We use multiple timeframes from 1 min to 1 hour, but always generate/update trade signals on one minute bars. Our exits are resting SL/TP OCO orders, so they of course get filled whenever they get filled, but we update them once per minute.
Wow... I'm shocked and happy to read you are using TA. There isn't much support on this board for TA trading. I'm not very smart with how algos work, so this may be a dumb question, but how does an algo analyze TA? Visually, we of course look for the standard things like higher lows or highs, double tops, reversals are key levels, some people use fib retracements, etc. Do your algos look at it the same way? I absolutely love this! I was never a fan of people saying you need to go back 10 years. I always thought that markets were fundamentally different back then, especially if we consider that 10 years ago, the algo trading was I'm sure much less. So why use data for analysis when things are done differently this time.