// FORWARD TEST

23 candidate rules, running on live bars

Our published signals come from one rule. We tested a much wider set — five timeframes and dozens of ways to close a position — and several did better in simulation than the rule we publish today. A backtest is not evidence, so those candidates are now running forward on live market data, and this page is their scoreboard while they do it.

Nothing here is published or sent to anyone. These rules read the same bars our scanner reads, decide independently, and record what they would have done. No subscriber receives them, they never touch the live track record, and no entry, stop or symbol from them appears anywhere — including on this page. If one earns its place, it changes what we publish, and we will say so here with the numbers before and after.

This is deliberately slow. Several of these rules work on weekly and monthly bars, so they take a bar a week or a bar a month, and a meaningful number of finished trades is months away rather than weeks. A page that sits still for a while is the honest outcome, not a broken one.

15340 open · 4500 resolved · running since 2026-08-24

AExit when momentum reaches a set level
Daily
open842
resolved540
win rate16%
median-4.65%
BTrailing stop, no fixed target
Daily
open891
resolved488
win rate7%
median-5.12%
CExit at the further trend line
Weekly
open123
resolved19
win rate5%
median-8.36%
DExit at the nearer trend line
Weekly
open107
resolved32
win rate50%
median-1.48%
EDifferent entry, exit at the further trend line
Weekly
open35
resolved12
win rate75%
median+0.96%
FExit at the further trend line
Monthly
open0
resolved0
win rate
median
GExit at the nearer trend line
Monthly
open0
resolved0
win rate
median
HDifferent entry (slower), exit at the further trend line
Weekly
open29
resolved12
win rate67%
median+0.76%
IDifferent entry, exit at the further trend line
Monthly
open47
resolved0
win rate
median
JRule A on the signal stream we actually publish
Daily
open211
resolved129
win rate19%
median-5.13%
KRule C on the signal stream we actually publish
Weekly
open36
resolved6
win rate33%
median-4.62%
LExit at the nearer trend line
Quarterly
open0
resolved0
win rate
median
MExit at the further trend line
Quarterly
open0
resolved0
win rate
median
NRule A on weekly bars
Weekly
open125
resolved18
win rate0%
median-8.45%
ORule A behind a wider stop
Daily
open873
resolved335
win rate20%
median-4.76%
PRule A on weekly bars behind a wider stop
Weekly
open130
resolved13
win rate0%
median-11.18%
QRule A on weekly bars behind a wider stop again
Weekly
open133
resolved10
win rate0%
median-11.18%
RA different entry: a setup that must form, fail and form again
Daily
open25
resolved22
win rate27%
median-3.36%
SRule B's exit on a different entry: price closing above its trend line
Daily
open1357
resolved93
win rate3%
median-5.78%
TRule S on weekly bars, behind a tighter trailing stop
Weekly
open286
resolved8
win rate0%
median-17.13%
URule T's exit on the momentum stream: the pair changes only the entry
Weekly
open80
resolved5
win rate0%
median-10.58%
VA weekly entry as momentum turns back up after a decline, behind a trailing stop
Weekly
open95
resolved0
win rate
median
WA daily entry as price climbs from below into a moving price band, behind a trailing stop
Daily
open63
resolved0
win rate
median

Win rate and median are long positions only, and every rate is computed from resolved positions alone — an open position counts for nothing until it closes. Short positions are recorded as a control and kept out of these numbers, because every backtested figure we have published is long-only.

Only 3809 of those 4500 resolved positions were recorded after the 1 September fix that let this test store the strategy our live signals use — the rest predate it and contain none of it.

A word on what the first weeks will show: the earliest records here are mostly crypto, commodities and foreign exchange, and our research population was US equities. So the first numbers on this page describe a different market from the simulation that produced these candidates. They will not match, and we would rather say that now than explain it later.

Letters A to G are the same rules that carry those letters in the write-up. The rest were not in that table and have no published backtest row: two are further entries, two re-run A and C on the exact signal stream our subscribers receive — the pair isolates the stream from the rule — and two cover the quarterly timeframe, which we scan and publish but never backtested.

Until one of these earns it, nothing changes: the signals we send come from the same rule as always, and there is no paid tier. Questions or suggestions are welcome — info@algosignals.finance.

Not financial advice. Simulated and experimental results carry no guarantee of future performance.