price_absorption

Real-time order-flow tooling for CME index futures (ES / MES / NQ / MNQ). The monitors read the live tape from Interactive Brokers, classify each trade as aggressive buying or selling, and fire audible alerts when the flow does something notable. There are also offline scripts for pulling historical bars and backtesting simple strategies against them.

Everything here is discretionary decision support — the tools read the tape and alert, they do not place orders.

What's in here

Live monitors (need a running TWS / IB Gateway)

Script	Detects	Idea
absorption_monitor.py	Absorption — heavy one-sided aggressive flow while price refuses to move.	Passive size is defending a level. Buy absorption (heavy selling, price holds) = support; sell absorption (heavy buying, price holds) = resistance. You supply the location by watching your chart.
absorption.py	Older absorption implementation, driven by config.py / run.py. Also logs post-signal MFE/MAE to absorption_analytics.txt.	Same concept, uses a max price-change threshold instead of a high-low range check.
momentum.py	Momentum — heavy one-sided flow with price moving the same direction.	The opposite of absorption: aggressors are winning, expect continuation.
tickstrike.py	Tick-rate spikes — trades-per-second bursting above baseline, regardless of delta or price move. Classifies the burst bullish / bearish / mixed from the buy/sell volume ratio.	Catches activity surges the delta-based monitors miss.

All monitors:

• Resolve the front-month contract automatically when --expiry is blank.
• Classify trades by comparing print price to the live bid/ask (at-ask = BUY, at-bid = SELL, mid = classified by tick direction).
• Compare a short detection window against a longer baseline window and trigger on a multiplier and an absolute floor (to suppress low-volume noise).
• Have a per-signal cooldown.
• Play a .wav from sounds/ on a signal, falling back to macOS system sounds (afplay) if the file is missing.

Offline scripts

Script	Purpose
historical_data.py	Downloads 1-second TRADES bars for a date range from IB in 30-min chunks and writes a *_1sec_RTH_*.csv. Handles IB's endDateTime timezone quirk (see comments in the file).
backtest.py	Replays your actual executed trades (Trades_*.csv) against historical 1-sec bars with a multi-rung ladder / stop / take-profit model. Time-of-day and daily-guardrail filters.
backtest_strategy.py	Self-contained RSI mean-reversion backtest on resampled 1-sec data, with slippage, ladder, cooldown, and trading-hours filters.

Setup

pip install ib_async pandas numpy

Requires Python 3.10+ (uses X | None syntax; developed on 3.14).

You need Interactive Brokers TWS or IB Gateway running and logged in, with the API enabled:

• TWS live: port 7496
• TWS paper: port 7497
• Gateway: port 4002

Live tape monitoring requires a CME real-time market-data subscription on the account.

Usage

Live monitors

# Absorption (recommended entry point), ES front month, defaults
python absorption_monitor.py

python absorption_monitor.py --symbol MES --min-delta 250   # less chatty
python momentum.py --symbol MES --points 3
python tickstrike.py --symbol ES --multiplier 3 --min-ticks 15

Common flags across monitors: --host, --port, --client-id, --symbol, --expiry (YYYYMM, blank = front month), --exchange, --window, --baseline. Run any script with --help for its full list.

Run each monitor with a distinct --client-id if you want several at once.

Config-driven runner

run.py reads config.py instead of CLI flags:

python run.py            # absorption (default)
python run.py momentum

Edit thresholds, symbol, and connection settings in config.py.

Historical data + backtests

python historical_data.py 20260701 20260730     # -> mnq_1sec_RTH_20260701_to_20260730.csv
python backtest_strategy.py                       # edit params at top of file
python backtest.py                                # needs a Trades_*.csv export

Backtest parameters live as constants at the top of each backtest script.

Notes

• sounds/ is empty in the repo — drop in buy_absorption.wav, sell_absorption.wav, etc., or let it fall back to system sounds.
• Alert sounds use afplay, so audio is macOS-only. Everything else is cross-platform.
• CSV timestamps from historical_data.py are in Central Time (exchange time).