We've found that slippage makes a huge difference to profitability. Slippage can be thought of as extra fees.
For bot flow (pdr_backend/trader): fees and slippage are naturally modeled in trader bots, because they're trading on real world data in real time. We've observed this in practice.
For simulation flow (pdr_backend/predictoor/approach3): it already models fees. But it does not model slippage.
Jaime's simulations do model slippage by bringing in historical + realtime orderbook data purchased from kaiko. By modeling it, and accurately, he was able to devise a trading strategy to account for it.
But this is not yet part of the approach3 simulation flow
And it uses data that opf had to purchase
In early Oct 2023, Trent met the founder of openmesh.network (Ashton) who shared that they have historical + realtime orderbook data, for free.
A hitch: the non-ccxt API will be different than ccxt. So it'll take effort to change things
Bonus: the non-ccxt may have Binance trading pairs available. If yes, this could be the cleanest solution to pdr-private#37 "US would-be predictoors can't see binance data from ccxt or binance api"
TODOs
[ ] Find free historical + realtime orderbook data. Starting point: play with openmesh.network
[ ] Update simulation flow (and maybe more) with new data source for status quo data
[ ] Update simulation flow to model slippage
[ ] (Bonus) update simulation flow where trader accounts for slippage, using Jaime's heuristic
[ ] (Bonus) update trader bot flow to account for slippage using the the heuristic ^
Background / motivation
We've found that slippage makes a huge difference to profitability. Slippage can be thought of as extra fees.
For bot flow (pdr_backend/trader): fees and slippage are naturally modeled in trader bots, because they're trading on real world data in real time. We've observed this in practice.
For simulation flow (pdr_backend/predictoor/approach3): it already models fees. But it does not model slippage.
Jaime's simulations do model slippage by bringing in historical + realtime orderbook data purchased from kaiko. By modeling it, and accurately, he was able to devise a trading strategy to account for it.
In early Oct 2023, Trent met the founder of openmesh.network (Ashton) who shared that they have historical + realtime orderbook data, for free.
TODOs