Absolutely. And even me saying what I said would give you a good idea for why or where I'm struggling. The thought that this drawdown should be avoided means that I am looking for the "perfect" entry, and this is of course difficult, and maybe an unrealistic endeavor. If instead a trader looks for a high probability that a trend will continue, then maybe the entry doesn't matter as much. But I focus on keeping losses low vs. where the profits are. The thing though is that most traders would assume the right thing to do is to exit for a loss, since you can always get back in. But this has its own challenges. First you took a loss, and then you're still looking for another entry. Clearly you can only do this if you have reasons to suspect that you go the direction right since fighting the trend is gonna kill you at some point. But its obvious your stats back up what you're doing. Darn. I was expecting though that you would say you're not really sure what the algo is doing or why, except that the minimum threshold of variables suggested to go long. Trading would really be much easier if you know you got the direction right. I'm sure there is nothing most of us could use directly from what you say. Its of course interesting to understand the thought process, because this can sometimes help a trader see things differently, but without understanding all the variables used, I think the secret is safe! Seeing as how you tried literally millions of combinations and variables to produce only several combinations that yielded positive results means none of us will stumble on this by accident! But I sure don't mind seeing more of your results as they come in!
Do you have a team of inhouse coders or do you contract out? The whole setup, how big your team and I imagine you'd require deep pockets to fund this project?
I get the same result in my local MC simulation in Excel with an expectancy of $180 per trade gross (excluding costs). However, only a 10 % drop in win rate changes the expected outcome dramatically with an expectancy of $60 per trade gross and some simulations going negative after 100 trades. If the win rate drops to 75 % your expectancy is $0 gross per trade, so effectively negative after costs. And that's pretty much my point. High win rate strategies with a negative R/R ratio really depends on maintaining that high win rate. Even a small drop in win % can effectively destroy your performance. So, can a 90 % win rate be maintained indefinitely? What if you compound and enter a losing streak after increase in size? Of course. My comments were more general, not towards this specific algo. So maybe they were misplaced in this thread. You may have lost me there, unless your point was that an algorithm isn't subject to emotions like manual traders, i.e., no difference between a 5 x 1 point drawdown versus 5 x 10 point drawdown. Mathematically, it's still subject to the same rules as every other system.
I have four part-time people I work with (who all have full-time jobs elsewhere), two developers and two Math/AI PhD's. I'm the only "full time" person on the strat/tech/trading side of things. Expenses were (relatively) minimal to get started, as we all worked for free for future profit-sharing and not for an hourly contract/wage. Of course we still had thousands a month in hard costs (Google Cloud infrastructure, physical servers, data feeds, software licenses...) during development. So really it was working for free AND paying for the privilege of doing so... The biggest expense was time, and not getting paid for it during development, working together over several years in the hope we were on to something. Turns out we were. Now, between sharing profits from performance-based income from our clients and trading in our own accounts, we're doing just fine, and the initial investment in time was obviously (in hindsight as the outcome is now known) worth it.
How do you attract these clients? Fundseeder? Norm Zadeh's competitions? etc? I wish I have a rich uncle that can bankroll me 100k for me stress test my algo properly. But my algo is not really scalable since it functions off stop orders so it's a liquidity chug. Maybe I can do 500k-1mil across multiple accounts before that DOM starts to act funny around openign range lol. What's the max scale you can use on your algo? Like in the couple of million?
Just shoulder-tapping and word of mouth, no advertising or anything public-facing. Once you have your first client that manages a large portfolio for others (which is the hardest part), and return an amazing ROI for them, word gets out quickly to other entities with money they're connected to. We chose S&P E-Mini futures contracts because of the insane volume. At any moment you can look at the orderbook and there's thousands of contracts at the bid/ask price (and those are just the resting orders you can see). We can place a $25M market order and not move the market one tick. And with limit/stop orders, I don't even know what our max scale is. If we ever get to where we move the market even one tick, we'll start implementing additional IB algo order types like Accumulate Distribute and Iceberg/Reserve orders.
Interesting. I trade the MNQ for now and hoping to transition to NQ once I get to a certain account threshold. I know s&p is very "thick" compared to the naz. But I just couldn't make my algo work properly on the MES/ES. Maybe because it's too liquid to have the extended move that I need, so it's very mean reversion type of a market. Maybe that's a sign that my algo is not robust enough? It's a modified ORB trading method. Anyway, thanks for the info.
Not at all? I spoke to another member here in private a few years back who ran an automated system on NQ. He said the same system didn't generate profits on ES. NQ and ES may seem similar enough, but the differences are noticeable when one starts taking a closer look.
ES trading range is like 1/3 or 1/4 of the NQ I think. When I backtested it, I have days where NQ would yield out tiny profit but ES would just literally kill me lol. I call ES the slow fat bastard. NQ is the sexy fast sprinter.
Right. I've added NQ to my repertoire now and may even replace ES completely. It simply moves more with more opportunities and better R/R ratios on average. Entering in the direction of momentum it's very easy to B/E your entry since it moves so fast or just collect an instant 5-10 point scalp if that's your thing. The same type of entry on ES will typically revert on you the instant you're filled. So, I can definitely see how these differences in micro-structure can make or break a strategy.