GECKOFINDINGS
The headline conclusions.
Every finding from Gecko's research, newest first - not the individual backtests behind them (see Insights for those), just the conclusions themselves: what held up, what didn't, and what's still an open question.
The hypothesis proved correct
The hypothesis was proven incorrect
Cautiously promising, but with caveats worth reading before trusting it
Reference point, not a hypothesis test
That verdict is relative - it says whether the result improved on whatever it was tested against, not whether it makes money on its own.
Tom Hougaard's Advanced School Run: A Real Edge at the Open - That CFD Costs Swallow Whole
Tom Hougaard's Advanced School Run brackets the DAX's 4th five-minute bar with a buy order above and a sell order below. Over 14 years of one-minute… Read the full finding →
An AI Read 10,000 Company Statements: Too Late for Swing Trades, Just in Time for the Open
Gecko had an AI read ten years of UK results and trading statements sceptically - the hard figures, the legally loaded 'ahead of / in line with /… Read the full finding →
When Directors Buy Their Own Shares: Gecko's First Edge from Outside the Price Chart
UK directors must announce every dealing in their own company's shares. Gecko read ten years of these notices - 33,000 of them - and found that large… Read the full finding →
Net result: £4,408.68
Short-Selling Disclosures: Heavily Shorted UK Shares Don't Reliably Fall
The FCA publishes every hedge-fund short position of 0.5% or more in a UK company. Research says heavily shorted shares underperform, so Gecko tested… Read the full finding →
Head and Shoulders, Both Ways Up: The Most Famous Chart Pattern, Done by the Book - and Still No Edge
Gecko built the head-and-shoulders exactly as Jack Schwager's textbook describes it - the bullish bottom (buy after a fall) and the bearish top… Read the full finding →
The Opening Price Principle: Mostly a Random Walk in Disguise
A trading workbook claims the day's opening price is near the high or low 70% of the time, and that you can trade away from it with the open as your… Read the full finding →
Bull Flags: Common, Easy to Spot - and Only Useful with Volume
The bull flag is one of the easiest chart patterns to define in code, and Gecko found nearly 2,000 of them. On their own they predicted nothing. The… Read the full finding →
Net result: −£5,817.23
Cup and Handle: Teaching Code to See a Chart Pattern
Can an algorithm spot the classic bullish cup and handle the way a trader's eye does - and once it can, does the pattern actually predict a rise?… Read the full finding →
Net result: −£361.66
Textbook Chart Patterns, by the Book: Lo, Mamaysky and Wang on UK Stocks
Gecko implemented the best-known academic definition of ten classic chart patterns - head and shoulders, double bottoms, triangles, rectangles - and… Read the full finding →
EVS Drift: The First Edge That Survives Honest Fills
A new strategy built from the lessons of TFM's failure: buy the morning after an exceptional volume spike, hold for 20 days. Tested with realistic… Read the full finding →
Net result: £10,134.12
The Forward Test Found the Flaw: TFM Only Worked With Perfect Fills
Four paper trades, four losses, each far bigger than planned. The cause: every backtest assumed orders fill exactly at their price, even when the… Read the full finding →
Net result: −£61,622.78
Shares vs CFDs: How Dropping Stamp Duty Brings TFM Back Into Profit
The same TFM strategy that loses £39,306 over ten years as UK share trading makes £17,966 when traded as CFDs. CFD overnight financing cost just… Read the full finding →
Net result: £17,966.28 (assumed perfect fills)
The VWAP Filter: Only Take Signals That Closed on the Right Side of the Day
A new entry filter for TFM - the signal day must close above its VWAP for a long (below for a short). On 115 stocks over ten years it lifts profit… Read the full finding →
Net result: £22,250.73 (did not include Stamp Duty; assumed perfect fills)
TFM's Other Exit Levers, Tested Against the RR 3.0 Baseline
Having settled on TFM's reward:risk ratio, the remaining exit-plan levers - target-zone selection, trade-expiry window, entry buffer - were swept one… Read the full finding →
Net result: £16,000.56 (did not include Stamp Duty; assumed perfect fills)
TFM's Best Result Yet: Stacking Three Validated Improvements
RR 3.0, second-nearest zone targeting, and a small-cap exclusion filter - each validated independently - combined into a single configuration and… Read the full finding →
Net result: £16,845.12 (did not include Stamp Duty; assumed perfect fills)
TFM's Reward:Risk Ladder, Properly Validated
After fixing overlapping trades, blind pending orders, leveraged position sizing and a permutation-test reconstruction bug, TFM's reward:risk sweep… Read the full finding →
Net result: £12,017.51 (did not include Stamp Duty; assumed perfect fills)
TFM Overtakes ORB: Gecko's Most Profitable Strategy Yet
Comparing TFM's tuned baseline against ORB's best-ever result and its flat full-scale result on the same 42-stock basket - TFM wins on profit factor… Read the full finding →
ORB at Scale: Wider Basket, Longer History, Both Directions — and the Edge Is Gone
Fixing a bug that had silently dropped every SHORT trade from every published ORB result, then scaling the strategy's baseline configuration up to… Read the full finding →
Net result: −£509,099.31 (did not include Stamp Duty; assumed perfect fills)
ORB on Banks vs the Broad Market: Real Edge, or Random Chance?
Rerunning every ORB variant tested so far against just the five UK banks, then using a permutation test to check whether the improvement there is a… Read the full finding →
Net result: £1,450.90 (did not include Stamp Duty; assumed perfect fills)
Overnight vs Intraday - Where does a stock's return actually happen?
Decomposing daily returns into the overnight and intraday segments to check whether ORB, an intraday-only strategy, is even fighting on the right… Read the full finding →