# I went quiet for a while. Here's what happened. I tested, 55 trading strategies properly, and almost all of them failed. — retained record

> Retained reddit record from reddit:algotrading.

Canonical: https://highsignal.app/data/records/a130297b8b55ba1c

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:algotrading)  published 2026-11-01 03:17 ingested 2026-11-02 02:42

[original source ↗](https://www.reddit.com/r/algotrading/comments/1xkssyv/i_went_quiet_for_a_while_heres_what_happened_i/)

## retained text



```
It's been a while since I posted, so here's the honest update.

\*\*The short version:\*\* I stopped building and started proving. Instead of adding more features to the bot, I spent the last stretch testing whether the ideas behind it actually work. Most don't. That's been the most useful thing I've done since starting AlgonikHQ.

\---

**Where I've been**

The forex bot has been running on a demo account for months. It looked \*close\*: some green weeks, some red weeks, hovering around break-even. The problem with "close" is that you can't tell if you're nearly there or just getting lucky.

So I built a proper testing setup and put every idea through it. The rules:

\- Rules first**.** Entry, exit and what counts as a pass are written down \*before\* I see any results. No moving the goalposts.  
\- **Real costs**Every trade pays the spread, overnight financing and a slippage allowance.  
\- **The flip test** Every signal also gets traded backwards. If the reverse does just as well, the "edge" is luck.  
\- **The random test** Same stops and exits, random entry times. A real signal has to beat 90% of random runs.  
\- **Unseen data.** Design on one period, check on another, and keep some data locked away for the final test.

About 55 variants. Tens of thousands of simulated trades. FX, stock indices, gold, oil, grains, bonds. 15-minute charts up to weekly. 2005 to now.

\---

**What I found**

1. My bot's original entry was picking direction at random. Traded backwards, it did the same. That one hurt, but it explained the months of "nearly there".

2. **Costs decide everything.** Short intraday trades pay more in spread than the edge is worth. Trades held for weeks pay more in overnight financing than they earn. The only cheap zone is holding for a few days and closing before the weekend.

3. **Rising markets fool backtests** One trend system looked brilliant on stock indices. Then random long entries made money too. It wasn't the signal. It was the market going up.

4. **One idea survived the flip and random test.** It's a momentum idea, in the same family I've always traded. It's statistically real. But it's small, and after costs there isn't much left. So it's not going live. It's being tracked live on paper instead, to see if it holds up on data nobody has seen.

5. **Knowing when to stop testing is part of the method.** Test enough variations and one will look amazing by pure luck. I've paused the pattern hunting for that reason.

\---

**What's running now**

\- The demo bot, with tighter risk rules, automatic failure alerts and a weekly report measured honestly in R (multiples of risk).  
\- A "shadow" tracker that logs the surviving signal every night and records how it plays out. No orders, just evidence.

Nothing goes near real money, or a prop-firm challenge, until it clears targets I've written down in advance: at least +0.1R per trade after all costs, a profit factor of 1.2 or more, 50+ trades, and drawdowns well inside prop-firm limits.

\---

**Why I'm sharing the failures**

Most trading content only shows the wins. I'd rather show the whole picture, including the bits that didn't work, because that's where the real lessons are. The goal hasn't changed. FIRE@45 is still the destination and the bot is still the vehicle. I'm just making sure the vehicle actually works before I put fuel in it.

The full research log (method, results table and lessons) is public here:  
[https://github.com/AlgonikHQ/algonik-research](https://github.com/AlgonikHQ/algonik-research)  
The exact rules and code stay private, but everything you need to see how I test, and what failed, is in there.

Monthly updates from here on, good or bad.

Not financial advice. Demo results only.

If you've been through something similar, or you know a cost or test I'm missing, I'd genuinely like to hear it.
```



document fetched 2026-11-02 00:38 · canonical https://reddit.com/r/algotrading/comments/1xkssyv/i_went_quiet_for_a_while_heres_what_happened_i
