Tessen
The truth about your trading strategy
Start new thread

Tessen - The truth about your trading strategy

byā€¢
Describe a trading strategy in plain English ā€” or paste Pine, Python, or MQL. Tessen builds it, grades it on 6 years of data with a real out-of-sample split, net of fees, then forward-tests it nightly in public. Most ideas fail. Knowing first is the product.

Add a comment

Replies

Best
Maker
šŸ“Œ
Hi PH ā€” founder here. I built Tessen after learning the expensive way. I ran an automated crypto bot with a 78% win rate and still bled money; when I finally backtested the strategy honestly over 6 years, it was flat. The win rate was hiding a 1:0.34 reward:risk. Nothing I'd read, bought, or copied had ever been measured properly. So Tessen is the measurement tool I wish had existed: ā€” Describe a strategy in plain English (or paste Pine / Python / MQL) and it becomes an exact, deterministic spec ā€” no vibes, every rule stated. ā€” It's graded on 6 years of data with a strict train/test split, net of real fees, across 8 assets, against 5 gates (expectancy, fees, drawdown, robustness, overfit). ā€” Passing grades don't end there: every graded strategy can be forward-tested nightly on fresh exchange data, and we publish the running record whether it flatters us or not. The part I'm proudest of is what fails. We graded 12 of the most famous strategies on the internet ā€” RSI dips, MACD, Golden Cross, Ichimoku, Bollinger both ways ā€” and 0 of 12 pass. We published every verdict with a permanent verify link. We gave 8 AI models an identical "design your best strategy" prompt, ported each answer rule-for-rule, and graded those too (the current leader sits at a B; the table is public). Grading is free and doesn't need an account. If you paste a strategy that can't be ported faithfully, it says so instead of guessing ā€” honesty is the whole moat. What it is NOT: financial advice, a profit promise, or a signal seller. A grade is a historical statistical measurement. Most strategies fail; that's the point of measuring. Happy to answer anything, including hostile methodology questions ā€” those are the fun ones.

The nightly public forward-test is a bold move, love that the grading is net of fees on a real out-of-sample split. Cuts through all the backtest hype most tools sell.

Maker

Ā Thank you ā€” the public scoreboard was honestly the scariest thing to ship. It updates nightly and can't be edited after the fact, so if our grades were flattering nonsense, the forward results would expose it within weeks. That pressure is kind of the point: , published win or lose.

Out of curiosity, what do you trade? If you run something through /grade and it fails somewhere that surprises you, I'd genuinely like to hear about it ā€” the surprising failures are where we learn the most.

Would love to see an option to export the equity curve and trade log as a CSV or PDF report so I can drop it straight into a journal or share with a co-pilot without copy-pasting screenshots. Also helps with post-mortems when a strategy fails the nightly check.

Maker

Ā You just caused a launch-day ship. The nightly forward-test trade log is now downloadable as CSV ā€” every recorded trade, win or lose, straight from the same table the scoreboard reads. Link at the top of (equity curve = cumulative (1 + net) per strategy, so it drops straight into a journal or spreadsheet).

Full trade-log + equity-curve export for your own graded strategies is the bigger lift ā€” the engine currently stores aggregates, so we need to start persisting per-trade data. Your comment just moved that up the roadmap. PDF reports too.

This is exactly the kind of feedback we launched for ā€” thank you.

the nightly forward tests in public are genuinely the part that sells me here, anyone can claim their strategy works on backtested data but showing it survive week after week on live data is a brutal honesty most platforms avoid

Maker

Ā This comment gets exactly why we built it this way ā€” thank you. The part that keeps us honest: the scoreboard can't forget. When a strategy that passed our gates starts bleeding on live data, its decay is right there in public, permanently ā€” we don't get to quietly delete the losers, and there already are some on the board. That discomfort is the product working.

Small addition from this morning: another commenter asked for the raw data, so the full forward-test trade log is now downloadable as CSV from ā€” every trade behind the scoreboard, win or lose, if you want to verify the honesty claim yourself rather than take it from a landing page.

the out-of-sample split being baked in upfront is such a smart move, basically removes the temptation to peek and tweak. honestly respect the team for putting the nightly forward tests in public too, that kind of transparency is rare

Maker

Ā Thank you, Elifnur! The upfront split was born from getting burned ourselves ā€” it's amazing how "just one more tweak" turns into curve-fitting when you can see the holdout data. So we made peeking structurally impossible: the split happens before grading, and the grade is what it is.

The public forward tests keep us honest too. Every strategy that passes goes on the scoreboard permanently ā€” including the ones that fall apart live. If we quietly deleted the losers, the wins would mean nothing. You can watch it update nightly at .

The out-of-sample split approach feels way more honest than backtests that conveniently curve-fit to the same data. Watching my ideas fail publicly is humbling but useful.

Maker

Ā That's exactly the trade, Enes ā€” the split makes curve-fitting structurally harder, but it also means you watch things fail in the open instead of quietly reworking until they pass. We eat the same medicine: every strategy we ship goes on the public forward-test scoreboard, wins and losses both, permanently.

Honestly the "humbling but useful" part is the whole point ā€” a grade you can't secretly retry is the only kind worth trusting.

Finally a tool that tells you upfront when your "genius" strategy is going to lose money. The public nightly forward testing is a nice touch, keeps everyone honest.

Maker

Ā Ha ā€” the quotes around "genius" are doing a lot of work, and we've all been there. That's exactly the moment it's built for: the out-of-sample split happens before grading, so it tells you the strategy loses before you've had time to talk yourself into it. And the nightly forward tests keep us honest too ā€” everything that passes stays on the public board, losers included. Appreciate you, Tayfun.