Signal Desk

The numbers a fund's risk team checks before allocating. Sharpe and Sortino measure return per unit of risk (above 1 is good); Value at Risk is the loss you'd expect to exceed on only 1 day in 20; beta shows how hard the coin swings when Bitcoin moves. Risk-free rate assumed 0%.

How the selected coin did in every month of every year. The bottom rows show each month's average and how often it finished up, which reveals seasonal patterns. The current month uses the live price and is still moving.

How it works: for each horizon, a logistic regression model learns from about 7 months of this coin's hourly candles, plus extra data with real history where your network can reach it: funding rates, Fear & Greed, the daily trend and, for altcoins, Bitcoin's own moves. Each horizon is trained with and without the extra data and keeps whichever tested better. It's trained on the older 70%, tested on the most recent 30% it never saw, and its percentages are recalibrated on that test so they reflect how often it was actually right. A model with no edge is pulled back toward 50/50. The 3-day and 7-day odds come from a separate model trained on years of daily candles (momentum over days to months, distance from the 20, 50 and 200-day averages, volatility, volume, Fear & Greed and, for altcoins, Bitcoin's week), tested the same way. Probabilities, not promises; short-term crypto moves are mostly noise.

Every hour, the direction odds' predictions for each coin you view (and every coin in the scanner) are saved in this browser. Once the hour, 4 hours, 12 hours, day, 3 days or week has passed, each prediction is checked against what price actually did. Unlike the backtests, nothing here uses past data: these are real predictions made in real time, so it's the most honest test there is. It builds up the more you use the page.

Watches for the conditions behind sharp turning points: panic selling into a capitulation low, or euphoria into a blow-off top. When pressure is stretched and price then reverses, a setup fires. Every past setup is backtested below so you can judge how much to trust it.

How it works: a setup arms when the pressure score reaches 50 (minor) or 65 (major), then fires on the first day within 5 days that price closes beyond the previous day's high (bottoms) or low (tops), confirming a reversal. Setups that fire during an open 20-day hold count as re-fires, not new trades. Use this as historical context, not instructions; even the best setups fail regularly.

Options prices tell you how far traders are paying to protect against or bet on moves. The ranges and odds here come from the whole "smile" of option prices for each expiry, not just one number. They are "risk-neutral": they include what traders pay for insurance, so they tend to overstate the chance of falls and assume no expected gain. Research and the live test above both point the same way: options are much better at predicting how big moves will be than which way they'll go. Sources: Bitcoin implied-volatility forecasting study, why option-implied odds differ from real-world odds.

Each cycle is rescaled to start at 1× so they can be compared directly (log scale). Hover or tap the chart to compare cycles at the same point. Past performance doesn't guarantee future results.

Targets assume this cycle repeats patterns from past cycles. Bitcoin's uses its gain from the cycle low; Ethereum's and Solana's use where their price against Bitcoin peaked in past cycles. Progress is measured on a log scale, the way cycle gains compound.

Past Bitcoin cycles

Every cycle so far has followed the same rhythm: a low roughly 1–1.5 years before a halving, a peak 1–1.5 years after it, then a crash of 75–85%.

Performance across timeframes

Returns over each period, using the live price. Strength across every column is what traders call momentum; a coin that's green short term but red long term is bouncing inside a downtrend.

How the coins move together

Correlation of daily returns over the last 90 days. 1.00 means they move in lockstep; near 0 means unrelated. High correlation means holding several of them gives less diversification than it looks.

Scroll to zoom, drag to pan, hover for details. The history uses only ingredients with real history, and each day only uses data available by that day's close. Live-only data (funding, options, order book) counts in today's signal counters but can't be in the history line.

Rules: each strategy starts with $10,000 and decides from each day's close whether to hold Bitcoin or cash for the next day, so it never trades on information it couldn't have had. Every switch costs 0.1% in fees. Cash earns nothing. To avoid flip-flopping, "avoid the caution zone" sells at 30 and only buys back at 40, "above 50" uses 45/55, and the 200-day rule needs a 2% move through the average. "No hindsight" builds the index with equal weights on every ingredient; "Evidence-weighted" uses the weights from the full history, which knows the future and will flatter itself. Past results don't guarantee future ones.

Log every trade and the stats below show whether your approach actually makes money. Most traders who keep a journal discover their losses come from a few repeat mistakes. Saved in this browser.

Important numbers are fetched from more than one place: one source is used and the other checks it, and if the main one is blocked on your network the backup takes over automatically. Small differences between sources are normal (they measure slightly differently); large ones are flagged.