I think if u combined it with footprint charts, and bookmap does show what they call collisions, this tells u about actual trades. But just resting orders on their own I don't think says much.
I used Bookmap for 6 months, had it running all day using integrated books for the top 6 crypto exchanges. Trying to see if it had an edge. The amazing thing it does is to see in real time what large players are up to, whales fighting each other around key support resistance areas. They try to push the market by bombing certain positions to move the stops. You can see a series of bombs being set at key positions in one go, which creates momentum.. then they wait on either side of the main support resistance line ready to catch the further market participants momentum bounce off or break through like the early arcade breakout game. So if you watch for this bombing raid you know something is afoot and can win small scalps whatever way things go, making guess about momentum. And so it helps to understand short term market behaviour for regular charts. For algo trading you could construct a 1 second timeframe strategy to detect the bombing raid then hedge both sides of the outcome. Because often you see the breakthroughs of the resistance line go on to breakthrough a whole load of others, and for some of those (that you hedge both ways) you could maintain a high 10 to 20X leverage on a 1-2% trade without stopping out. Possibly even further.. I had a whole thing worked out. But the main barrier was I had better long timeframe strats and there is no long term data I can backtest on. So am waiting for processing capability of these large volume data sets to become easier and for next generation of ai to run backtests on strats with it.
u dont need next gen ai to run machine learning algos on order book data. You do need to be faster to analyse and send orders to exchange
Faster than what ? You do need some type of ai if you want an automated trading system that can backtest correlation ideas on such huge volume datasets that are visually orientated. And you need next gen ai IF you want ai itself to carry out re-enforcement learning and testing on the visual type data bookmap has. Because its a lot of data for which you have to tell it to play a breakout type hedging game. And current generation Ai hallucinates and doesnt have proper self monitoring before it outputs decisions. You can used older ai of course but it will not be a straightforward process. And obviously you dont need ai, if you just want to do manual trading. Or implement simple ideas
my x2 cents worth is that I paper traded bitcoin with cumulative order flow on Bookmap and it kinda worked. As well I then paper traded fakeouts in a direction of the main trend and that kinda worked.
I tried it but couldnt get consistency. It only seemed to work during bull conditions, but then wouldnt work. And it depended which order book you integrated. Did you integrate all the biggest order books ? Because crypto is not centralised so orderflow from one broker exchange represents only fraction of the action. As all order books from each exchange have different delta, and different support/resistance heatmaps
I only had Binance data-feed. My test wasn't too clever, I didn't do any code or such, because free Bookmap doesn't let one save the data. I was mostly looking for convergence and divergence of delta at peaks. I was just eyeballing it to asses is it worth me saving the data and writing the code for the full blow back-test. But after a month and about 20 - 30 trades approach scraped some profit. But, what is really fascinating one gets to see high volume levels every $5.0. Basically on bar charts one is guessing levels, but with Bookmap you can see which levels are real and how strong they are. That must be worth some alpha.