Frequently asked
questions.

Ensitics is an esports prediction platform. Every day the model goes through the tournament schedule across esports disciplines, such as Counter‑Strike 2, League of Legends, Dota 2, Valorant, Mobile Legends, Overwatch, Rocket League, Rainbow 6, Honor of Kings, and publishes the matches where it sees an edge. A pick is two things: the side and a confidence level. Simple, easy to use.

What Ensitics isn’t: a stats database like HLTV or Liquipedia, and not a place where you can wager. You get the conclusion rather than the table you’d have to build it from.

A reasonable question to ask in this category — it’s full of Telegram channels and random websites that quietly delete the picks that lost. Three things you can check before paying anything here:

The thing worth being suspicious of, here or anywhere else, is a service that shows you its wins and not its losses.

No. We don’t take bets, hold deposits or pay out winnings, and we aren’t affiliated with an operator that does. Ensitics publishes an opinion about a match; what you do with it happens somewhere else entirely. That also means we have no interest in you betting more often — a subscription costs the same whether you act on 15+ picks a day or one a week.

A subscription to the daily pick feed. That’s the whole business. No affiliate deals with bookmakers, no VIP tiers, no upsell to a private channel. There’s a referral programme for existing users, and that’s it.

1,618 of 2,210 graded picks have won since launch. It’s 73.2%, as of September 12, 2026. Everything is based on the historical database legacy (10+ years) mixed with the actual stats.

Two caveats worth having before that number means anything. First, the win rate is not profit: a feed of heavy favourites can win 80% of the time and still lose money. Second, the two algorithms behave very differently, so a blended win rate hides more than it shows.

HC stands for High Confidence. Matches where the model is unusually sure and the market broadly agrees. Shorter odds, higher strike rate, less drama.

VS is Value Spotter. The opposite: matches where our number and the bookmaker’s number disagree. These lose more often. They also produced all five of our biggest wins so far: O2 Blast at 6.20 (Overwatch, 5 June), Enterprise Esports at 5.30 (Valorant, 7 August), SYGaming at 4.72 (Honor of Kings, 24 June), Gen.G Esports at 4.29 (League of Legends, 27 July), G2 Esports at 3.95 (Valorant, 22 August).

Mixing them without knowing which is which is the most common way to misread the feed.

Ensitics has a huge database of more than 12.1 million rows of data. This historical legacy makes the algorithm work.

The pipeline runs the same loop daily: scan every tournament across all the disciplines, pull the match‑level history for both teams (such as recent series, opponent quality, map and draft splits, roster state, event context), then filter down to only the matches that clear the threshold of one of the two algorithms. Most matches don’t. On a typical day the model looks at hundreds and publishes about 15. Might be more, rarely less.

The output is deliberately narrow: a side, a confidence band, and minimum odds. We don’t publish a probability for every fixture, because a number attached to a match we have no read on is worse than silence.

Partly, and it’s worth being precise about which parts. Models are good at what esports produces in bulk: thousands of graded series, map‑level splits, head‑to-head history, opponent strength, form curves. They’re poor at what decides a surprising share of matches: a stand‑in announced an hour before the game, a patch landing mid‑tournament, a roster playing its final series before disbanding, a team already qualified.

So the honest framing is calibration, not prophecy. A model calling a match at 70% should lose roughly three of those in ten. If it doesn’t lose them, the number was wrong in the other direction.

Published claims in this space range from about 60% to “89% verified”, and the spread says more about counting methods than about model quality. A tool that only picks heavy favourites can post a very high win rate while losing money; one that only reports its best sample can post anything at all.

Yes. 592 losses out of 2,210 predictions is roughly one in 4, and they don’t arrive evenly. The best stretch we’ve had was 20 picks in a row without a loss, between 12 July and 15 July 2026. Runs in the other direction happen too, which is the argument for sizing bets against confidence rather than treating every published pick as equivalent.

Yes. One pick a day, permanently, no card required to register. It’s the same model output as the paid feed, just less of it. The point is that you can grade us yourself for a month before spending anything.

$8.90 per month for the full feed: 15+ picks a day across all the titles, plus the complete prediction history. Cancel whenever — there’s no annual lock‑in and no tier above it.

Not to read the picks. Plenty of subscribers use the feed to check their own reads or to follow a scene they can’t watch live. To act on a pick you’d need an account with a licensed operator in your jurisdiction, which is a separate thing entirely. We don’t place bets and don’t send you to a specific bookmaker — that’s an important part of the value.

Yes, on iOS. Currently it has 4.6 on the App Store. The web version at ensitics.io has the same feed and the full history.

A prediction is a stated view on how a competitive match will go. Usually which team wins, sometimes the map score or a specific in‑game market. It is not the same thing as a bet, and not the same thing as a tip. A prediction can be entirely correct about a team’s strength and still be a bad bet if the price already reflects it, which is why any prediction worth reading arrives with a price attached.

The signals that carry weight, roughly in order below.

Doing this by hand for one match is an evening’s work. That’s the gap the model fills. Not better judgement necessarily, just the same checks run across the whole schedule every day.

A bet where your estimate of the probability is higher than the price implies. If you think a team wins 50% of the time and the odds are 2.40, that’s value, even though you expect to lose half of those bets. It’s the whole basis of the Value Spotter algorithm, and it’s counterintuitive enough that we wrote it up separately: what a value bet is and how to find one.

Some of them are, yes. The problem isn’t that they’re free. It’s that free picks are usually published without a record, so you can’t tell a good source from a lucky week. Before following anyone, paid or not, look for a graded history that includes losses.

Counter‑Strike 2, generally. The map pool is fixed, the veto is public, and the game changes slowly enough that a season of data stays relevant. League of Legends and Dota 2 are harder in patch weeks and much easier in the settled middle of a split or season. Mobile Legends sits in between, with deep regional leagues but thinner cross‑region data.

Format matters as much as title. Any Bo1 in any game is the hardest thing on the schedule to call, and it’s where most upsets in our loss column come from.

It depends on where you are, and it’s changed a lot recently. For example, several US states regulate esports wagering separately from traditional sports betting, and some don’t permit it at all. We can’t answer it for your jurisdiction and wouldn’t try. Ensitics itself doesn’t accept wagers or hold funds, so nothing on this site is a bet; whether you can act on a pick with a licensed operator is a question for your local regulator. The service is 21+.