Performance

Measured on data it never saw.

Every number on this page comes from trades the system decided without knowing the outcome, with trading costs taken off. Where it did badly, it says so. Backtest, paper and live results are never added together.

Backtest

Walk-forward tests over up to 19 years of history, then a final holdout year no development decision was ever based on.

Paper (forward)

Every signal shown since launch, logged permanently and settled by its own exit rules. Can't be edited or cherry-picked.

Live trading

None. Radarline never places trades or connects to a broker, so there is no live trading record - and we won't show one.

The short version

What the testing found

Tested on 65 markets across FX, metals, energy, grains, indices, bonds, US stocks, ETFs and crypto.

Predicting the next candle from indicators is close to a coin flip on every timeframe. The model's probabilities beat a constant guess, but only modestly.
Most of the 'edge' in rising markets is simply being long. Every strategy here is compared with just buying, with the same exits - and two beat it: buying pullbacks in uptrends, and breakouts.
No short-selling strategy survived testing, so the system doesn't show SELL signals. If one ever passes, it will.
Choosing markets by their own past record - going forward, never in hindsight - cut drawdowns by about two thirds.
Charts under 1 hour lost money after costs in every strategy tested, so there are no 15-minute signals.
A 90% win rate is easy to get and loses money: tiny targets, big losses. Nothing tested earned money at 85%+ wins.

Backtest · final holdout

The year it was never allowed to see

Frozen first, then run once on the most recent period. New is Radarline Signals; current v3 is the previous engine's trade signals under the same realistic execution; always long buys every candle with v3's exits - the bar any system should clear. Results in R: 1R is the amount risked on a trade.

Daily charts · 24 Sept 2025 to now (1.0 years)

System	Trades	Win rate	Avg per trade	Profit factor	Max drawdown	Worst streak	t-stat
New - Radarline Signals	275	62.9%	+0.030R	1.13	14.0R	7	0.6
Current v3	821	62.0%	+0.067R	1.17	34.9R	10	2.2
Always long	1861	60.0%	+0.036R	1.09	68.6R	14	1.8
New, before market selection	503	52.1%	-0.052R	0.85	48.7R	12	-1.0

A $10,000 account risking 1% per trade, max 5 positions at once, daily loss limit and drawdown stop.

New Current v3

New: took 118 of 275 signals (the risk rules skipped the rest: at most 5 open, daily loss limit, losing-streak pause, drawdown stop), ended -7.3%, worst drop 11.0%.

Current v3: took 162 of 821 signals (the risk rules skipped the rest: at most 5 open, daily loss limit, losing-streak pause, drawdown stop), ended +11.1%, worst drop 12.7%.

4 hour charts · 24 Sept 2025 to now (1.0 years)

System	Trades	Win rate	Avg per trade	Profit factor	Max drawdown	Worst streak	t-stat
New - Radarline Signals	378	50.0%	+0.133R	1.37	24.6R	11	1.5
Current v3	3377	59.2%	-0.001R	1.00	82.8R	11	-0.1
Always long	10107	58.2%	-0.019R	0.96	336.0R	15	-2.2
New, before market selection	1791	48.9%	+0.019R	1.05	102.5R	13	0.5

A $10,000 account risking 1% per trade, max 5 positions at once, daily loss limit and drawdown stop.

New Current v3

New: took 132 of 378 signals (the risk rules skipped the rest: at most 5 open, daily loss limit, losing-streak pause, drawdown stop), ended +40.7%, worst drop 20.0% - hit the 20% drawdown stop and stopped trading.

Current v3: took 419 of 3377 signals (the risk rules skipped the rest: at most 5 open, daily loss limit, losing-streak pause, drawdown stop), ended -14.2%, worst drop 21.5% - hit the 20% drawdown stop and stopped trading.

1 hour charts · 24 Mar 2026 to now (0.5 years)

System	Trades	Win rate	Avg per trade	Profit factor	Max drawdown	Worst streak	t-stat
New - Radarline Signals	215	25.1%	+0.133R	1.17	64.1R	33	0.4
Current v3	6211	56.6%	-0.107R	0.78	680.8R	29	-9.4
Always long	22521	57.0%	-0.094R	0.80	2170.7R	21	-15.9
New, before market selection	1320	26.1%	+0.035R	1.05	238.3R	29	0.4

A $10,000 account risking 1% per trade, max 5 positions at once, daily loss limit and drawdown stop.

New Current v3

New: took 29 of 215 signals (the risk rules skipped the rest: at most 5 open, daily loss limit, losing-streak pause, drawdown stop), ended -19.6%, worst drop 20.7% - hit the 20% drawdown stop and stopped trading.

Current v3: took 245 of 6211 signals (the risk rules skipped the rest: at most 5 open, daily loss limit, losing-streak pause, drawdown stop), ended -12.0%, worst drop 22.2% - hit the 20% drawdown stop and stopped trading.

Read this honestly

On daily charts the previous engine did better in this particular year - its constant long bias suited a rising market. The new system's dip-buying did well; its breakouts had a bad year.
On 4-hour charts the new system clearly beat both, mostly thanks to choosing markets by their record.
One year is not enough to be statistically sure (t-statistics under 2). The case rests on the long walk-forward tests below and on this holdout pointing the same way.
1-hour breakouts win only about 1 trade in 4 and can lose 30 in a row. Size accordingly, or leave them.

Paper · forward record

Since launch

Every BUY / STRONG BUY the site has shown, settled by the strategy's own exits. This is the record that matters most over time.

No settled paper trades yet. Signals are logged from the moment they appear (last market scan 26/09/2026, 11:03:53 am); results fill in here as each trade reaches its exit. Nothing is back-filled.

Probabilities

Does 65% mean 65%?

Each signal's chance of a win comes from a model calibrated on separate data. The test: of all trades it gave about 65%, did about 65% win? Hollow dots have too few trades to judge.

Daily · holdout

Brier 0.206 vs 0.233 for a constant guess (lower is better)

4 hour · holdout

Brier 0.229 vs 0.250 for a constant guess (lower is better)

1 hour · holdout

Brier 0.186 vs 0.188 for a constant guess (lower is better)

Backtest · strategy research

What was tested, and what survived

Daily charts, 19 years of development data, costs included and then doubled and tripled. A strategy only counts if it beats its control - the same exits with no entry signal - and survives the correction for how many ideas were tried.

Strategy	Trades	Win rate	Avg per trade	At 3x costs	Longs / shorts	Real after 30 trials?
tsmom	13902	49.7%	+0.020R	+0.009R	+0.080R / -0.074R	78%
pullback	11885	64.7%	+0.021R	+0.006R	+0.038R / -0.005R	100%
pullback long-only	6970	66.7%	+0.038R	+0.024R	+0.038R / -	100%
breakout	4428	29.8%	+0.166R	+0.145R	+0.440R / -0.169R	98%
breakout long-only	2446	34.1%	+0.438R	+0.415R	+0.438R / -	100%
Control: long every uptrend candle, pullback exits	56395	59.4%	+0.007R	-0.008R	+0.007R / -	100%
Control: long every candle, breakout exits	11481	25.9%	+0.174R	+0.154R	+0.174R / -	100%
always long, monthly	13890	53.2%	+0.071R	+0.061R	+0.071R / -	100%

Adaptive

Where it has an edge - and where it doesn't

Each strategy, market and timeframe keeps its own record. A market only gets signals while its record clears zero; everywhere else the answer is NO TRADE. Markets with a qualifying record today:

Market type	Daily	4 hour	1 hour
ags	2 of 3	0 of 3	0 of 3
bond	0 of 2	0 of 2	0 of 2
crypto	3 of 8	0 of 8	5 of 8
energy	2 of 3	0 of 3	1 of 3
etf	5 of 7	4 of 7	5 of 7
fx cross	1 of 5	0 of 5	0 of 5
fx major	2 of 7	0 of 7	0 of 7
index	7 of 9	4 of 9	0 of 9
metal	4 of 7	5 of 7	5 of 7
stock	11 of 14	9 of 14	2 of 14

The 90% question

Can it win 90% of the time?

Yes - and it loses money doing it. Long trades on every daily candle, 1.5 ATR stop, with smaller and smaller targets:

Target	Win rate	Avg win	Avg loss	Worst trade	Avg per trade
0.1R	89.5%	+0.086R	-1.028R	-6.323R	-0.031R
0.25R	79.9%	+0.236R	-1.029R	-6.323R	-0.018R
0.5R	67.6%	+0.485R	-1.028R	-6.323R	-0.005R
0.75R	59.1%	+0.732R	-1.024R	-7.242R	+0.014R
1R	52.5%	+0.969R	-1.018R	-7.242R	+0.026R
1.5R	44.4%	+1.370R	-1.002R	-5.929R	+0.052R
2R	40.6%	+1.630R	-0.992R	-6.436R	+0.072R
3R	38.4%	+1.847R	-0.987R	-5.693R	+0.100R

Radarline optimises for expected profit per trade and smaller drawdowns, not an impressive win rate. Its pullback signals win about 2 in 3; its breakouts about 1 in 3 but with much bigger winners.

Method

How it was tested

Data

65 markets from TradingView (up to 20 years daily, 2-10 years 4-hour, 7 months-3 years hourly) plus 6 years of hourly and 15-minute crypto from Binance.

No look-ahead

Every input uses closed candles only - tested by recomputing on truncated history. Trades enter on the next candle's open; stops are checked candle by candle, gap fills included, and a candle touching both stop and target counts as the stop.

Costs

Spread, commission and slippage per market type (1.5 bp for major FX up to 16 bp for crypto, round trip), then everything re-run at 2x and 3x.

Walk-forward

Models and market choices were only ever learned from trades that had finished before the period being tested, with a gap so no trade straddles the boundary.

Holdout

The most recent year (six months for hourly) was fenced off and run exactly once, after every decision was frozen.

Honest limits

Today's markets were chosen with hindsight (survivorship bias); most years tested were rising markets; news wasn't backtested (there's no archive to test it on), so it's shown only as a risk warning.

Past and backtested results do not predict future results. General information only, not financial advice - see the General Advice Warning.