Liquidity Heatmap
sweep-reclaim rule · Sep 18
Its mechanical sweep-reclaim rule showed no standalone edge in backtest; on the chart it is read as context.
Now: In use as context
Read the log →The MGE research pipeline
Most ideas don't survive the pipeline. That is the point of it. Choose a stage to see what is there.
Every step is checked against a random-entry control and out-of-sample data, and we correct our own numbers in public when we find a mistake.
What counts as evidence?
Passing the backtest is not the finish line. It is permission to test further.
Fail any one of the four and the result is usually luck, costs or a bug, not an edge.
Research status
Last updated Sep 26, 2026. This shows how far each idea has got, not how well it did. It is not a result.
- FlowEdge06 Refined Sep 23
- CSC / AB Trader05 Forward test, Sim
- MarketCoach V3.4.105 Forward test
- EMA50 Chandelier05 Forward test
- NYLowSweep05 Checking fills
- Exhaustion Coil04 Backtest
- Liquidity Heatmap04 No standalone edge
- Bar Structure04 No standalone edge
What failed
We publish what fails. Most ideas don’t survive the pipeline. That is the point of it.
IB 25/75
retracement rule · Sep 15
The indicator stays, the strategy was closed.
Now: Indicator in use as a map
Read the log →MarketCoach
relaxed config · Sep 12
A passing walk-forward test was resting on roll gaps.
Now: Original config, forward test
Read the log →Kraken VolSqueeze
breakout bot · Sep 2
First real test, no edge demonstrated.
Now: Paper bot stopped
Read the log →Bar Structure
bar sequences as entry rules
A reading aid, not a signal: our backtests of the bar sequences it names found no standalone edge as entry rules.
Read the log →PlayCall
bar-sequence plan
A template, not a signal: our own test of the pattern lost a little on average on every time frame, so it is read by hand.
Read the log →Deviation Reversal
session VWAP bands
Its mechanical backtest was negative; it stays on the chart only as a picture of how far price has stretched.
See it on a chart →CSC / AB Trader
forward testing on simulation only
Our mechanical backtest of the coded rules did not find an edge after costs. A handful of trades proves nothing either way.
Read the dev log →From the weekly notes: what failed
- Our own test had two bugs. Here is how we caught them.A control that beats the real strategy is a warning sign, not a discovery.
- Zero wins out of thirty-three looked like a disaster. It was a bug in the counting.The scariest number on the screen that day was not a losing trade. It was a bug that had never lost a real one.
- A passing walk-forward test that failed once the data was correctedBoth configurations got worse once the data was fixed. What we changed, and what stayed.
No live results are claimed for anything still in use; forward tests are how we find out.
Built on real charts
Our tools on real NQ charts, the good calls and the bad ones. No mock-ups.
A signal firing shows the machinery works; it does not show an edge.
Backtest ≠ proof
Historical test vs. current state. A mechanical backtest tests one fixed version of a rule. It can find no edge while the same tool stays useful in a trader's hands, read their own way. We publish both sides.
| Model | Mechanical backtest | Now | Date |
|---|---|---|---|
| Liquidity Heatmapsweep-reclaim rule | No standalone edge | In use as context | Sep 18 |
| IB 25/75retracement rule | 22 signals in ~7 yrs | Indicator in use as a map | Sep 15 |
| MarketCoachrelaxed config | Pass rested on roll gaps | Original config, forward test | Sep 12 |
| Kraken VolSqueezebreakout bot | Negative over 109 trades | Paper bot stopped | Sep 2 |
No live results are claimed for anything still in use; forward tests are how we find out. Read the write-ups →
A strategy in action
CSC / AB Trader, our automated breakout strategy, recorded in playback on a simulation account.
The MGE Quant stack
How an idea moves from raw data to a tested rule.
01
Data
- Market structure
- Liquidity
- Volume
- Price action
02
Research
- Hypothesis
- Model development
- Statistical testing
03
Models
- FlowEdge
- MarketCoach
- CSC / AB
- Exhaustion Coil
04
Validation
- Backtest
- Random control
- Out-of-sample
- Forward test (Sim)
05
Execution
- Entry
- Risk
- Stops and targets
- Trade management
The MGE Framework work in progress
The parts we keep testing together: Structure · Liquidity · Flow · Context · Entry · Risk · Execution. The bar is high on purpose: no tool has cleared all four evidence tests yet. We publish every result, pass or fail.
See what passed and what failed →Prop-firm calculator
Check an account against its drawdown, consistency, payout and trading-day rules before you take the trade.
Account$50,000
Drawdown$2,500
Consistency30%
Latest research
Why we build in public
Research should be documented, not marketed. We show the process: the experiments, the failures and the lessons.
Our mission →Plain-language glossary
The words used on this site, in everyday terms.
- R
- One unit of risk: the amount you are prepared to lose on a single trade. A trade that makes twice that amount is +2R; one that loses it is −1R.
- Edge
- A repeatable advantage: a rule that does better than chance would, after costs. Most ideas turn out to have none, which is why we test them.
- Backtest
- Running a fixed set of rules over old data to see what they would have done. It tests one version of a rule, and passing it is not proof.
- Out-of-sample
- Data the rules never saw while they were being built. A result that only works on the data it was tuned on is usually luck.
- Forward test
- Running the rules on new data as it arrives, on a simulation account, to find out whether the backtest holds up.
- Random-entry control
- The same test with entries taken at random, using the same stops, targets and timing. If the rule cannot beat it, it has shown nothing beyond chance.
Educational and research content only. Not investment advice. No trade signals are sold or given. Trading futures involves substantial risk of loss. Backtested and simulated results are hypothetical and do not represent actual trading.