LoL match predictions: how to read them, how to write your own

“LoL match predictions” as a search query returns thousands of articles that all look the same. Team A preview. Team B preview. Recent form. “We are going with Team A for this one.” No price discussion. No probability. No closing line analysis. Useless.

This article is about how to read the predictions that are good, how to spot the ones that are noise, and how to write your own.

Anatomy of a useful prediction

A useful prediction includes:

A probability estimate, not just a pick.
A price at which the bet has value.
The reasoning, with specific inputs.
A clear “no bet” threshold.

Everything else is padding.

If you read a prediction that says “Team A wins 2-1,” no probability, no price, no reasoning beyond recent form, you are reading entertainment, not analysis. Sometimes entertainment is fine. Just do not bet on it.

Red flags in LoL prediction articles

You can tell a bad prediction in 30 seconds. Signs:

“Lock of the day” language.
Odds quoted once at the end, no probability discussion.
Reliance on head-to-head stats without sample size discussion.
Sentences like “they have momentum after last week” without defining momentum.
No mention of patch, draft, or roster stability.
The writer does not publish their track record. Ever.

The last one is the most telling. Anyone claiming edge should be willing to show a spreadsheet. The unwillingness to publish a track record is a confession.

What a good prediction actually sounds like

Here is a template:

Gen.G vs T1, BO3, LCK Summer Week 5, patch 14.18. Model says Gen.G 53%, T1 47%. Closing line is Gen.G 1.90 / T1 2.00 (no-vig: 51.3% / 48.7%). Model disagrees with market by 1.7%. No bet.

If Gen.G drifts to 2.05 (no-vig ~48%), the model disagrees by 5 percent. Take Gen.G. Watch the close.

Inputs: Gen.G is +850 GD15 over last 20 maps vs T1 +620. Gen.G top priority strong on current patch (Aatrox S-tier, they are 5-1 on it). T1 jungler returning from stand-in stretch, re-integration week 1. Blue-side for Gen.G game 1. No other news.

Three things make that prediction useful:

It is a decision, not a pick. “No bet” is a valid answer. In fact, it is the most common correct answer.
It is conditional. A price change changes the decision.
It exposes the inputs. You can argue with them.

Writing your own before you bet

This is a habit worth building. Before any bet, write a prediction in exactly this format:

Match. Teams, format, event, patch.
My probability. What do I think the real chance of each outcome is?
Market no-vig. What is the market pricing, stripped of vig?
Gap. Is my number more than 3 percent different from the market?
Inputs. Three to five sentences on why my number is different.
Decision. Bet Team X at price Y. Or no bet.

The whole thing takes five minutes per match. It forces you to confront gaps between your opinion and the market, which is the only reason a bet makes sense.

If you cannot fill in the “inputs” section with specifics, you are betting on feelings. Feelings are the most expensive input in sports betting.

Common prediction mistakes (from reading 500 of them)

Overweighting recent single results

“Team A just beat Team B 2-0 last week, I am going with Team A again at 1.75.” That 2-0 is one data point. It could flip next time. Head to heads over 5 or fewer series are noise.

Ignoring the patch

People pick a BO3 winner without looking at the current patch. If the meta just shifted, last split’s form is stale. A model built on patch 14.14 data is not a model for patch 14.18.

Ignoring roster news

A team that changed a starter in the last 14 days is not the same team. The market usually has this priced. Prediction articles often do not. Read the actual roster, not the team name.

”Due for a bounce”

A team on a three-match loss streak is not due. They might be better or worse than their record suggests. “Due” is not a real input.

Using KDA as a predictor

KDA is a retrospective stat. It describes what happened. It predicts less than gold diff at 15 does and much less than draft priority does. Use it as flavor, not as input.

When the market moves before your analysis

Sometimes you do the work, arrive at a prediction, and then the line moves in the direction you were going to bet before you place it. What now?

If the line moved 2 percent or less, your edge shrank but might still exist. Recompute the no-vig price. If you still have 3 percent edge, bet.
If the line moved 3 to 5 percent, your edge is probably gone. Look for inferior prices at other books.
If the line moved 5 percent or more, sharp money agrees with you, and they beat you to it. The value is gone. Pass.

The goal is not to be right. The goal is to get good prices. A right prediction at a bad price loses money over time.

The “no pick is a pick” principle

Most of the good prediction writers I know place bets on 20 to 30 percent of matches they analyze. The rest, no bet.

Compare that to a typical tipster, who “predicts” every match on the schedule. That is the tell. The market is not that soft. If you have an opinion on every match, your opinions are not doing much work.

A useful prediction service, blog, or account tells you what not to bet as often as it tells you what to bet.

Where our predictions sit

We publish predictions only when our model disagrees with the market by at least 3 percent at the prices available. On a typical week, that is 2 to 5 matches out of 15 to 20 on the schedule.

The live predictions page is at /predictions. The full model writeup, with assumptions and hit rate over 280 tracked wagers, is in the prediction methodology article.

If you want picks without the math, there are a thousand sites that will give them to you. This is not one of them, and it is better for both of us.