New strategies are now in forward test
A short update on what we are building, and why.
Where the live product stands
Our published signals are performing well. Of 270 tracked signals, 232 have resolved and 149 of them reached their target — a 64% win rate, with every signal published before its outcome was known. Wins and losses are both on the track record, and nothing is ever removed from it.
We scan 1,773 instruments across 26 markets every day, and that part of the system is doing its job.
What we tested
Over the past weeks we ran a much wider study: five timeframes, several entry methods, and dozens of ways of closing a position — hundreds of thousands of simulated trades in total. The question was simple: are there rules that do better than the one we publish today?
There are. Here are the strongest, with names and parameters left out on purpose — they are still being validated.
| Candidate | Timeframe | Trades tested | Win rate | Avg per trade | Typical hold | Annualised |
|---|---|---|---|---|---|---|
| A | Daily | 64,900 | 58.7% | +1.31% | 25 bars | 13.3% |
| B | Daily | 64,900 | 39.8% | +1.16% | 19 bars | 15.1% |
| C | Weekly | 10,100 | 67.5% | +3.52% | 16 bars | 11.7% |
| D | Weekly | 10,400 | 80.2% | +1.41% | 8 bars | 9.3% |
| E | Weekly | 5,600 | 67.9% | +9.41% | 13 bars | 38.8% |
| F | Monthly | 3,300 | 75.7% | +13.16% | 7 bars | 23.2% |
| G | Monthly | 3,900 | 81.4% | +9.52% | 5 bars | 22.4% |
Three things worth saying plainly. These are long positions only, before commission and slippage. The annualised column is the per-trade result scaled by how often the rule trades — one position at a time, not compounded, and not a portfolio return. And candidate E was measured on data that played no part in choosing it — the check that separates a real result from a well-fitted one.
What happens now
Nine of these rules are running in forward test as of this week. They read the same live bars as our scanner, decide independently, and record what they would have done — in a ledger that publishes nothing and sends nothing. No subscriber receives anything from them yet.
This takes time by design. Several of them work on weekly and monthly bars, so a meaningful number of completed trades is months away, not weeks. We would rather wait than publish a rule that looked good for six weeks.
Since publishing this we have opened that ledger to the public: the forward test page shows every candidate, how many positions each has open, how many have resolved and what they returned — aggregates only, no symbols and no levels, because these rules still publish nothing.
If they hold up on live data, we plan to open them to subscribers in a few months. Until then, nothing changes: the signals you receive are produced by the same rule as always, and there is no paid tier.
Where this is going
The longer-term goal is the reason we are being this careful. A rule precise enough to be tested mechanically is a rule that can be executed mechanically — and our aim is to turn these strategies into automated trading through established brokers, so a validated setup can be acted on without manual work.
That is the direction we are building toward. We will keep publishing what we find on the way, including the results that do not go our way.
Talk to us
Questions, comments or suggestions are welcome — send them to info@algosignals.finance. If there is something you want measured, tell us; a few of the tests above exist because someone asked.
Not financial advice.
⚠️ Not financial advice.