What would happen if your strategy could see, and respond to, the order flow after it entered?

General Discussions
1

To me, it makes sense that the strat should “know” when to push its target out while potentially tightening its stop.

For instance, if a bunch of buying comes in the moment your strat goes long, wouldn’t the forces of supply and demand push the price higher?

I’ve tested this on 63,350 trades from different unprofitable ES strats (Mean Reversion, Scalping, Momentum, Breakout, etc.), and there seems to be enough of an edge to justify having the strat update its stops and targets according to post-entry order-flow behavior.

Here are my findings:

01 Mean reversion
Profit Factor: 0.7882 → 0.8835 (+12.1%)

02 Range rotation
Profit Factor: 0.6645 → 0.8390 (+26.3%)

03 Trend & momentum
Profit Factor: 0.8034 → 0.9246 (+15.1%)

04 Pullback & retracement
Profit Factor: 0.7038 → 0.8054 (+14.4%)

05 Breakout
Profit Factor: 0.7195 → 0.7946 (+10.4%)

06 Scalping & microstructure
Profit Factor: 0.5010 → 0.5130 (+2.4%)

07 Failed-break reversal
Profit Factor: 0.6043 → 0.7321 (+21.1%)

I’m seeking additional sample data.

If you have an ES strat and want me to see whether post-entry order flow can help improve your returns, I have a proposal for ya.

First off, you keep your strat and its logic private. I do not want to know your secret sauce. What I could use is just your trade list of entries and exits.

From there, I can run my algos over the duration of each trade and see whether post-entry adaptive trade management, based on order flow, can goose the strat’s returns.

LMK if you have a trade list for an automated ES strat that you might want me to pump through these algos.

The offer is open to the first 10 quants/traders who provide a trade list, the bar type and bar size used, and the time zone used for the trade-list timestamps.

Please send:

  • Unedited NinjaTrader trade-list CSV
  • Bar type and size, such as 5 Minute, 2,000 Tick, or 9 Range
  • NinjaTrader time zone

Hopefully, we can grow together.

Note: The current focus is on ES. I will be expanding to other markets once the dust settles on this.

2

GTFOH! and go pump your scarcity tactics on twitter.

1 Like
3

Didn’t mean to offend you. Feel free to leave me alone.

4

well before I will give you my sauce by giving you my trade list of entries and exits. I’m curious what you are using to test this on 63,350 trades on , maybe u are just using a simple system that is not good enough, maybe if u tell me what u are using I can guide you towards a better one, I don’t need your system settings I just need to know what Ur using, lol

5

I seem to have poorly explained myself.

I do not want secret sauce. I cannot reverse engineer your recipe from just entries and exits.

Is this understood?

If you think I can reverse engineer your secret sauce, can you shed light on how?

I am an order flow guy. I find this stuff fascinating.

I used off-the-shelf strats that I found. I built some trade management algos on top of order flow.

The point is not what system was it run on.

The point is, is post-entry adaptive trade management, predicated on order flow algos, something that might be viable?

From my findings, I think it needs more research.

Hence, the call for anonymous data - which are trade lists that cannot reveal ‘secret sauce’.

The concept of post trade adaptive management is something that I thought would be universally interesting.

I tried to set up a process where I could do the work with no risk or cost to anyone who might also be interested in this.

6

if you don’t mind me asking the terminal or system, software u using to get this trading information , to test this on 63,350 trades on

7

For this exercise, I used Ninja’s Strategy Analyzer. The strats are C#. I do not tune/optimize them.

I have routines that look at the entry date time and run alternate exits based on the level of buying or selling as determined by order flow algos.

The concept is pretty straight-forward. If the strat is long and there is buying in order flow, push the target out.

My current dev environment is Codex and I leave it on Sol 5/.6 Extra High.

I have also used QuantConnect, SierraChart, TradeStation and a bunch of other stuff.

I do long to ‘roll my own’ platform so I can do hard-tuning for this type of development/research/operations.

8

P.S. I wrote my first order flow algo in December of 2003, so I have over 20 years of work in this field.

9

I’m so confused. What do you mean by “for this exercise”? That phrase honestly made my brain buffer for a second.

The moment I read that, it felt like I became a test subject instead of someone having a real conversation. I thought this discussion was getting interesting and serious but calling it an “exercise” makes it sound like it isn’t.

To me, an exercise is something used for testing, evaluating, or training. It gives the impression that the model is still being experimented on or that the conversation isn’t being treated as genuine. That completely changed how I interpreted everything we had been discussing.

I was actually invested in the conversation, and then that one phrase pulled me out of it. If this is meant to be a real discussion, and the details on the systems u use, with in a 20-year background yeah there much missing there,

I admire your positivity and your drive to move forward quickly, but meaningful progress takes time. Every skill is built one step at a time, and trying to skip those steps usually creates bigger obstacles later.

The best results come from learning the fundamentals first, building a solid foundation, and then advancing gradually. Stay patient, trust the process, and keep moving forward steady progress will always outperform rushing ahead.

10

It is an exercise.

I am trying to validate that post-entry order flow trade management can consistently provide alpha.

Ours is a world of R&D.

I consider R&D a series of exercises. We could call them sprints, tasks, whatever.

11

I hope that people take this idea, their agentic coding LLM, some data, and run with it.

If folks want to collaborate with me, I’m here.