TPR Radar · the rules
How Radar works, exactly.
Every rule Radar follows, written down. Nothing here is a black box: if you have the candles, you can check every alert yourself. These rules are also what each alert is graded on.
Candles and RSI
Radar reads closed candles only. A candle counts once it has closed, so an alert never repaints: what was posted stays true to the chart it was posted on.
RSI is RSI(14) on closing prices with Wilder smoothing, the standard definition most charting platforms use.
Pivots
A pivot low is a candle whose low is the lowest of the 3 candles on each side of it. A pivot high is a candle whose high is the highest of the 3 candles on each side.
A pivot is only confirmed once the 3 candles after it have closed. So an alert built on two pivots fires 3 candles after the second pivot, never sooner. That delay is the price of never repainting.
RSI divergence
Two pivots of the same kind, 5 to 60 candles apart.
- Bullish: the second pivot low is lower than the first, but RSI at the second is higher than at the first. Price pushed lower with less force.
- Bearish: the second pivot high is higher than the first, but RSI at the second is lower than at the first. Price pushed higher with less force.
Watch: for a bullish alert, the highest high between the two pivots; for a bearish one, the lowest low between them.
Wrong if: price trades through the second pivot (below the second low for a bullish alert, above the second high for a bearish one).
Timeframes: crypto on 1H, 4H and daily; stocks on 1H and daily, regular trading hours only.
Overbought and oversold divergence
The same divergence, with one more condition: RSI at the first pivot was below 30 for a bullish alert, or above 70 for a bearish one. The move into the first pivot was stretched, and the second could not match it.
Watch and wrong-if levels are the same as for an RSI divergence. Crypto only.
Monday range sweeps
Monday's high and low are taken from its 1H candles, on UTC days. For stocks, "Monday" is the first trading day of the week.
From Tuesday, the first 1H candle that trades beyond Monday's high or low starts a sweep. Radar then looks at that candle and the next 5, six candles in all:
- Reclaim: if any of the six closes back inside Monday's range, the sweep failed and Radar fades it. Watch: Monday's midpoint. Wrong if: price trades beyond the sweep's furthest point.
- Hold: if all six close outside the range, the break held and Radar follows it. Wrong if: price trades back through the Monday level. Crypto only.
At most one sweep alert per side per week.
Funding extremes
Crypto only. Each hourly funding print is compared with the coin's own prints over the last 30 days.
- A print above 95% of them is bearish: longs are crowded and paying for it.
- A print below 95% of them is bullish: shorts are crowded.
Ties never count: a print equal to past prints does not push it over the line. Wrong if: price trades beyond the 24-hour high for a bearish alert, or the 24-hour low for a bullish one.
A coin needs at least 24 days of funding history before it can alert, so funding alerts switch on around 21 November 2026. At most one per side per coin in any 24 hours, on the 15 coins of Radar's funding board.
What Radar watches
Crypto: BTC, ETH and SOL, plus the next 25 coins by market cap, using Bybit perpetual futures.
Stocks: SPY, QQQ, IWM, NVDA, TSLA, AAPL, MSFT, META, AMZN, GOOGL, AMD, COIN, MSTR, HOOD.
How alerts are scored
R is the distance from the alert's price to its wrong-if level. An alert worked if price moved 1R in the alert's direction before it touched the wrong-if level. If both happen inside the same candle, it counts as not worked: we do not get to guess the order. An alert faded out if neither happened within its window:
| Market | Timeframe | Window |
|---|---|---|
| Crypto | 1H | 24 hours |
| Crypto | 4H | 3 days |
| Crypto | Daily | 10 days |
| Stocks | 1H | 7 trading hours |
| Stocks | Daily | 10 trading days |
An alert whose wrong-if level is closer than 0.2% or further than 15% from price is not posted: too tight to mean anything, or too wide to be useful.
The random baseline. For every alert, Radar also takes 5 random entries on the same timeframe, the same side and the same risk size, and scores them by exactly the same rules. The Sunday scorecard puts each alert type's hit rate beside that baseline. Beating random is the bar; if a type does not clear it, the scorecard says so.
Cooldowns. One alert per coin, per direction, per family: 8 hours on 1H, 24 hours on 4H, 3 days on daily. The two divergence types count as one family, so a coin does not get both for the same move.
What we tested and dropped
Some ideas did not make it. Here they are, so nobody wonders.
- Multi-timeframe combos, the same setup on two timeframes at once: too rare to matter, and no better than random when they did fire.
- Weekly and monthly open touches: the same as random.
- Stocks on 4H: overnight gaps break the candles, so stocks stay on 1H and daily.
- Monday holds on stocks: dropped; stocks get reclaims only.
The backtest, in plain words
The alerts beat random picks by a few points. That is real, and it is also not a strategy on its own: after fees, a few points is not enough to trade blind. What Radar does well is show where to look. What you do there is up to you.
If you want actual trades, with the entry, the stop and the risk, those are the traders' calls in The Room, and every one of them is on the public track record, losses included.
Back to TPR Radar · Everything else that is free
18+ only. Radar alerts are context, not calls or advice. Not financial advice. Trading carries a real risk of loss; backtested and past results do not predict future ones.