The Lab

What we tested

BTC Spot vs Leverage (SvL) v4.5 is a TradingView indicator that sets two things against each other:

• Spot pressure: is the coin being bought or sold on spot exchanges (Binance, Coinbase and Kraken)? It counts each 5-minute candle's volume as buying if the candle closed up and as selling if it closed down. (A rough cousin of trade-by-trade rules such as Lee and Ready's, 1991.)
• Leverage pressure: are futures traders crowding one way? It reads funding (Binance, Bybit, OKX), made stronger while open interest grows.

From the two it names six moments. Each one points the price a way:

It was made for Bitcoin on the 4-hour chart.

How we tested it

• We rebuilt it from its Pine source, line by line, using only public data that goes back two years: Binance and Coinbase 5-minute candles, Binance and Bybit funding, and Bybit open interest. Kraken keeps only 2.5 days of candles, OKX keeps 3 months of funding and Binance keeps 30 days of open interest, so those three were left out.
  • Pine (Pine Script) is TradingView's own language for chart indicators. An indicator's Pine source is its code, which anyone can read on TradingView. For example, ta.sma(close, 20) is Pine for the average of the last 20 closes.
• The window: 11 October 2024 to 10 October 2026, split into two years.
• Each signal was judged by the price 1, 3, 6, 12 and 24 bars later, in the direction the signal names, after paying a market order in and out (0.11% in all).
• Against chance: the same number of random bars, drawn 400 times. A signal only "beats chance" if its average is above 95% of the random draws.
• Five charts: 1, 2, 4, 8 and 12 hours. A real effect usually shows, a little weaker, on the charts next to the one it was made for.
• Three ways to count spot pressure: as built (candle colour, Binance plus Coinbase), Binance only by candle colour, and Binance's real record of who bought and who sold.

That makes 5 charts × 3 ways × 6 signals × 5 horizons = 450 tests.

The maths, step by step

One test, worked through: S? on the 8-hour chart, with real buying and selling, judged 6 bars (48 hours) later.

1. The signal's result. S? fired 26 times. Each is a sell signal, so a fall in price counts as a gain. If the price fell 1%, the result is 1% − 0.11% fees = +0.89%. The average over the 26 was +1.43%, and 69% of them gained (the hit rate).
2. Chance's results. Pick 26 bars at random, sell at each, and average the same way. Do it 400 times. That gives 400 averages from entries with no skill at all.
3. The two lines. Sort the 400 from worst to best.
   • The 20th from the bottom (5% of 400) is the 5% line: −1.11%.
   • The 20th from the top is the 95% line: +0.84%.
   • Random picks land between the two lines 9 times in 10.
4. The verdict. +1.43% is above the 95% line, so this test beats chance. Under the 5% line it would be worse than chance.

Why luck alone gives about 5% on each side: a signal with no skill is just one more random pick. A random pick lands above the 95% line 1 time in 20. So of 450 tests with no skill at all, about 450 × 5% = 22 would beat chance anyway, and about 22 would do worse.

Why the count is a lean, not proof: the tests overlap. S? at 3, 6 and 12 bars uses the same 26 signals. If those 26 happened to come before falls, all three horizons pass together. That is one lucky draw read three times, not three separate wins. So "40 beat chance where luck gives 22" counts some luck more than once.

What we found

1. On the 4-hour chart there was no reliable edge. Most signals lost after fees at most horizons. The buy signals (A, L?) did worse than random at times. D and S? came out ahead at exactly 6 bars (one day) but not at 3 or 12. A single good horizon among its neighbours looks like luck.
2. Counting real buyers and sellers didn't help on 4 hours. Swapping the candle colour for Binance's real record showed nothing there.
3. The faint signs were on the slower charts, and on the sell side. On 8 and 12 hours, with real buying and selling, S?, D and flush beat chance at several horizons. But there were only 7 to 58 signals each, and much of it came in the second year.
4. Overall, 40 of 450 tests beat chance. By luck alone you would expect about 22 (5%). Another 14 did worse than chance. The tests overlap heavily (the same signals, a few bars apart), so this is a lean towards the short side, not a finding.

Every signal as a trade. Each one is entered at its bar's close and left 6 bars later, after fees:

Three things to be careful of:

• S? is part of D. Every S? is also a D on the same bar, so the two aren't separate evidence.
• One sell-off carries the flush results. On 2 June 2026 Bitcoin fell from about 69,500 to 62,000. That one move gave 10.7 of flush's 18.4 points on the 12-hour chart.
• The 12-hour strength is mostly the second year. In the first year, D and S? there were a coin flip.

The one we followed up: S? on the 8-hour chart held up in both years. We then tested it on data it hadn't seen (Bitcoin's earlier years, ETH and SOL), and it did not hold. See The SvL on other charts and coins.

What this means

• The indicator is honestly built. It avoids repainting, and it tells apart moves led by spot buyers from moves led by leverage.
  • Repainting is when an indicator changes a past signal after the fact, because it used a price that wasn't known yet at that moment. For example, a signal shown at 10:00 that could only be worked out once the candle closed at 14:00.
• On the chart it was made for, its signals did no better than chance over these two years.
• The faint signs sat on the slower charts, outside its design, and did not hold on new data.
• Its thresholds were set by hand, and no test on unseen data comes with it. A test like this one is the check to run before trading on any indicator.

Limits

• Bitcoin only, and only two years.
• No Kraken spot data, no OKX funding and no Binance open interest, as explained above.
• Prices are Binance spot closes. There is no slippage and no funding cost on the trades, and wicks inside a 5-minute candle aren't seen.
  • Slippage is the gap between the price you expect and the price a market order fills at: the order takes whatever the order book offers. For example, with half a coin offered at 100 and the rest at 100.20, a market buy of one coin averages 100.10, 0.1% of slippage.
  • A wick is the thin line above or below a candle's body: how far the price went during the candle before coming back. For example, a candle that opens at 100, dips to 95 and closes at 99 has a wick down to 95.
• An indicator alone can't trade: these are signals, judged by where the price went next.

Update, 27 October 2026: where the idea comes from

Added after this page was published. The text above is left as it was.

• Spot against leverage. The idea that a rise led by spot buyers is sturdier than one pushed by borrowed money is widely held among crypto traders. We know of no paper that first stated it, so we name no one. The indicator puts the idea into rules. The rules are its author's.
• Counting volume as buying or selling by the candle's colour goes back to Joseph Granville's on-balance volume (1963). A day's whole volume is added if the price closed up and taken away if it closed down.
  • For example: up on 100, down on 40, up on 60 gives 100 − 40 + 60 = +120, read as buying.
  • The catch: a candle that closes up by a cent counts all of its volume as buying. A real trade record splits each candle into what buyers and sellers took. This page tested that record too.
• Funding comes with the perpetual future, a futures contract that never expires. Robert Shiller proposed perpetual futures in 1993; the first crypto ones opened in 2016. To keep its price near spot, the crowded side pays the other every few hours.
  • For example: funding of 0.01% every 8 hours on a $10,000 long costs $1 each time, paid to the shorts.
  • The catch: a crowd can stay crowded for weeks. Funding says who is leaning, not when they will be forced out.
• Judging signals against random bars is a Monte Carlo test (Metropolis and Ulam, 1949). Why every test needs one is on our theory page "Hard-to-beat markets, and why every test needs a random twin".
• Theirs and ours. The indicator, its six signals and its thresholds are its author's. Granville's counting, Lee and Ready's, Shiller's perpetual future and the Monte Carlo method are credited below. The rebuild, the five charts, the three ways of counting and the random-bar test are ours.

Credits

• SvL v4.5 (BTC Spot vs Leverage), credited at the top of this page, published on TradingView and shown here with permission. The indicator, its signals and its rules are theirs; we rebuilt and tested it.
• Charles M. C. Lee and Mark J. Ready, "Inferring Trade Direction from Intraday Data", Journal of Finance, 1991: classifying trades as buys or sells.
• Joseph E. Granville, Granville's New Key to Stock Market Profits, Prentice-Hall, 1963: on-balance volume.
• Robert J. Shiller, "Measuring Asset Values for Cash Settlement in Derivative Markets: Hedonic Repeated Measures Indices and Perpetual Futures", Journal of Finance, 1993: the perpetual future.
• Nicholas Metropolis and Stanisław Ulam, "The Monte Carlo Method", Journal of the American Statistical Association, 1949: drawing random comparisons many times.
• The replay and the random-bar test are our own design.
• Data: the exchanges' public candles, funding, open interest and taker records named on this page.
• PerpGremlin and Claude, A TradingView indicator, rebuilt and replayed on two years of Bitcoin, TerminalGremlin, 2026: the tests, the code and the words.