The Lab: find your edge before the market does

Signal leaderboard

The top-3 signal bot watches the top 3 here, kept up to date after every update: in each market, whichever of them fires first is its trade. Every trade since it started, wins and losses (a live-price simulation). See its record →

How it's ranked, and why. The site's scan tries hundreds of thousands of signal combinations every few minutes. Try that many and some will look great on the markets they were picked from by luck alone, the way one coin out of thousands can land heads ten times in a row. A signal's record on markets that came after it was found or added can't be picked that way, so that is what the leaderboard ranks by: $10 the first time it fires in each new market, at the asking price with the fee, from $100.

• Vetted (held up on new markets): 100 or more fires after it was found or added, and ahead on them by more than luck explains: its returns at least 3.5 standard errors above what breaking even would give at the prices it paid. The board looks again after every market, so the bar is set for that many looks: in a simulation of the board itself, a signal that only breaks even after the fee gets vetted in its first month about 1 time in 100, one making +10% a bet about two times in three. With dozens being tested an occasional lucky one can still get through. These come first, best first.
• Still being tested: everything else, ranked by its record since it was found or added (it says how far along the bar it is); a short record can still be luck.
• A Lab signal that is behind after 100 new-market fires, or down 40% or more after 30 of them, comes off. A template or Lab signal that, after 2 days on the board, has fired in under 3% of the markets since it was added comes off too: it hardly ever applies.
• Behind right now (greyed, at the bottom): the built-ins and templates that are behind, over the last 500 markets or on their record since they were added (and a Lab signal brought back by hand, in its first days). They are still tested after every market and move back up as soon as they pull ahead.

Under each name is its record over the last 500 settled markets as well. For a Lab signal that includes the markets it was found on, so treat it as a starting point, not proof. Lab signals join only when their luck check on windows the scan never saw reads "worth watching" or "unlikely to be luck", they fire in at least 40% of markets, are no lottery signal, and hold up walking forward (their levels set again from earlier stretches keep half or more of the return they were picked on, and are ahead in at least 2 of 3 stretches they never saw), and beat simply buying their side every market at the same prices on the windows they never saw, by 1 standard error or more (since Oct 31: a find that only rode BTC's trend, such as Down-only finds during a drop, makes money until the trend turns and then comes off; this keeps most of them out); Lock counts from its current version (Oct 25), the templates from when they joined the board. An arrow shows each signal's move since the last market. Tap any of them to load it into the simulator.

Hide the leaderboard

Members' signal board

Signals our Lifetime members made themselves, tested the same way as the leaderboard: over its last 500 markets, $10 the first time it fires in a market, at the asking price with the fee, its record counting only the markets after it was entered. One that gets vetted (100 or more such fires, ahead by more than luck explains) joins the leaderboard above, credited to the member who made it.

Paper tests

Scan for the best signals

The scanner tries signals made of one, two or three conditions on the page's numbers (RSI, the crowd, the model, the gap, the indicators and the rest), for as long as you choose, and lists the ones whose side did best when taken at the market's own asking price, $1 a window, the fee included. Trying thousands of signals on the same windows always turns up some that look excellent by luck, so the older two thirds of the windows choose the list and the newest third, which the scan never saw, shows which ones held up. Then run any of them in the simulator second by second, and put it on your chart.

Strategy simulator

Indicator leaderboard

These are the single indicators, each on its own (RSI, the crowd, whale orders and the rest). The signals on the signal leaderboard, the templates and the scanner's results are built from combinations of these indicators and conditions, so they are a different thing: a signal can do well where none of its indicators does on its own.

Over the moments 10, 5, 2 and 1 minute before each close: which way each indicator pointed, how often that side won, and what the market was already pricing for it. Only a side that wins more often than its price is telling you something new.