How SkinBetHub Picks CS2 Bets and Scores the Record

Most tip channels explain their picks with words like "form", "feel" or "inside info". Those are hard to check. This page explains where SkinBetHub CS2 picks actually come from, the rules they are posted under, how the record is scored, and why you should be skeptical of any short record, including ours.

Where the picks come from

Picks come from SkinBetHub's proprietary signal model, built on prediction-market data. Prediction markets such as Polymarket and Kalshi let people trade on match outcomes, and their prices move as money comes in. That data is the raw material. When the model flags a CS2 match as qualifying, it becomes a pick.

We do not publish the exact inputs or thresholds. They are the product, so they stay private. That is a fair thing to be wary of, which is why the record is built so you can judge it without knowing the method:

• Timestamped picks. Each pick appears on the live board before the match starts, so nothing can be added once the result is known.
• Posted odds. Every pick carries the price it was given at, so you can see whether it had value, not just whether it won.
• Flat units. Every pick is graded at 1 unit, so stake sizing cannot flatter the record.
• Forward-only ledger. The record starts from the first published pick, with no back-filled history, and every win, loss and void stays on /results.
• Sample-size math. A short record proves little either way. The sample-size section below shows how many picks it takes.
• Closing line value. If picks are consistently posted at better odds than the market closes at, that is evidence of value long before the win-loss record settles down.

No method is immune to cold runs, and past results on any market do not guarantee future ones. That is exactly why our own record is scored in public, from the first pick forward. You do not have to trust the model. You can check the ledger.

The posting rules

These are the same rules shown on the live board:

• Posted before the match. A pick appears with its odds before the match starts. Nothing is added after a result is known.
• Forward-only. The record counts picks from the day they were first published. There is no back-filled or simulated history.
• Flat 1 unit. Every pick is scored at the same stake, so there is no way to quietly put "max units" on the ones that won.
• Nothing is deleted. Every win, loss and void stays on /results.

Posting the odds matters as much as posting the pick. "Team A to win" tells you nothing about value. A favorite at 1.30 and the same favorite at 1.70 are very different bets, and only the price tells you which one the pick was.

When nothing qualifies, there is no pick

There is no daily quota. If the model finds no qualifying CS2 match, nothing is posted. Some days have several picks, some have none.

This is deliberate. Channels that post three "locks" every day are usually filling a schedule, not finding value. A skip costs you nothing, while a forced pick at a bad price costs you money over time.

How the results record is scored

The record uses the same definitions as our methodology page:

• Settled pick: a pick whose match has finished and been graded as a win or a loss.
• Void: a cancelled or voided pick. It stays visible but counts as neither a win nor a loss.
• Hit rate: wins divided by settled picks. Voids are excluded.
• Units: a win adds the posted decimal odds minus 1. A loss subtracts 1.
• ROI: total units won or lost divided by the number of settled picks.
• Break-even hit rate: 1 divided by the average posted odds. It is roughly the hit rate a record has to beat to show a profit.

Here is a small worked example with four picks:

• Settled picks: 3 (the void is excluded).
• Hit rate: 2 ÷ 3 = 66.7%.
• Units: 0.60 - 1.00 + 0.45 = +0.05.
• ROI: 0.05 ÷ 3 = +1.7%.
• Average settled odds: (1.60 + 1.80 + 1.45) ÷ 3 = 1.62, so break-even is 1 ÷ 1.62 = about 61.9%.

Notice how much the price matters. A 66.7% hit rate sounds strong, but at these odds it only beats break-even by about five points, and the profit is a twentieth of a unit. You can run the same numbers on any record with the ledger ROI calculator.

Why a small sample cannot prove an edge

Every betting record mixes two things: whatever skill is there, and luck. Over a few dozen picks, luck dominates. A confidence interval is the honest way to show this. It is the range the true ROI probably sits in, given what has been observed so far. A 95% interval means that if you repeated the exercise many times, about 95 of 100 such ranges would contain the true value.

Here is the arithmetic for a realistic case: picks at an average price of 1.50 with a true ROI of +3%.

1. A +3% ROI at 1.50 means a win rate of 1.03 ÷ 1.50 = 68.7%. Break-even at 1.50 is 66.7%, so the edge is two percentage points of hit rate.
2. Each pick returns either +0.50 or -1.00 units. The spread of that result (the standard deviation) is 1.50 × √(0.687 × 0.313), which is about 0.70 units per pick.
3. The uncertainty in the average shrinks with the square root of the number of picks: 0.70 ÷ √n. A 95% interval is about 1.96 times that on each side.
4. For the interval to clear zero, 1.96 × 0.70 ÷ √n has to be smaller than 0.03. That gives √n above 45.5, so n above roughly 2,070 picks.

After 100 picks, an observed +3% ROI is consistent with anything from about -10.6% to +16.6%. After 500 picks the range is still about -3.1% to +9.1%, which includes zero. Only around 2,000 picks does it start to exclude zero, and even that is the minimum: if the true edge is +3%, there is still about an even chance that an unlucky run leaves the observed number short of that line. Being reasonably sure of seeing it (80% of the time) takes about 4,200 picks.

So when any tipster, us included, shows a few hundred picks with a nice ROI, the honest reading is "not disproven yet", not "proven". Bigger edges show up sooner, and higher average odds make the result noisier, so the exact count changes. The direction does not.

One way to learn faster is to look at prices instead of results. If picks are consistently taken at better odds than the market closes at, that is evidence of value long before the win-loss record settles down. Our guide to closing line value in CS2 explains how to measure it.

How to read our record

• Look at the number of settled picks first, then ROI. Treat ROI on a small sample as a rough signal only.
• Compare the hit rate with the break-even hit rate, not with 50%.
• Check that losses and voids are visible next to the wins. They should be.
• Watch open picks on the live board before their matches start, then check they land on /results unchanged.

If you follow any picks, from us or anyone else, keep stakes small and flat, and never bet money you need. A process can be sound and still lose for months.