Why do I have winning and losing streaks in chess?

• A win raises your rating, so the next pairing is near that new number and the opponent is stronger. A loss lowers it, and Chess.com says you are then matched with slightly weaker players. A winning streak does not add a hidden modifier on top of that.
• Chess.com Live search starts within ±25 rating points and widens if nobody is waiting. By default, players under 2500 in bullet or blitz, or under 2000 in rapid, are only paired within ±200. You are paired with someone seeking the same time control.
• Chess.com does not describe matching by playing style. Its matchmaking article also says openings are not considered.
• Chess.com's help page says it uses the Glicko system, including a rating deviation. Lichess says it uses Glicko-2. Those are not the same sentence, so this page does not call Chess.com's system Glicko-2.
• Fatigue and tilt stretch a losing run. A stop rule, written before you play, is the practical limit. It is not a promise that your rating will rise.
• Judge a slump from the rating range over many games in one time control, not from a single day.

Chess.com's matchmaking article says this directly in its myth table: a winning streak does not make games harder by a hidden rule. Matchmaking uses your rating. Wins raise that rating, which pairs you with stronger opponents. Losses lower it, which pairs you with slightly weaker players, until the results even out. There is no separate "shadow rating."

That is why a bunch of wins and then a bunch of losses can be the system working. You are not parked at one fixed opponent strength. Each result moves the band of people you meet. Around your current rating, some stretches will be mostly wins and some mostly losses.

What changes the size of a rating move

Chess.com's ratings article says a rating does not move by the same amount after every game. Three factors change the size of the move:

• The rating gap. Beating someone much higher raises you more than beating someone similar or lower. Losing to someone lower drops you more than losing to someone higher.
• Confidence in your rating. High confidence means smaller changes. Low confidence means bigger ones. A new player, or someone returning after a break, moves more, because the rating deviation has gone up.
• Confidence in the opponent's rating. Chess.com says that if it is less confident in their rating, your change could be more significant. It also gives a specific case that goes the other way: if you lose to a much lower-rated player who has a high rating deviation (they have not played in a long time), your rating will not change much, because their number may not be accurate.

Chess.com says it uses the Glicko rating system, and that rating deviation is the confidence measure. It does not say it uses Glicko-2, and it does not publish a volatility number. This page does not add either claim.

On Chess.com, a brand-new time class does not start from zero. If you have already played other Live time classes, the new one starts at or near the closest one you have (rapid uses blitz before bullet, in their example). If you have played no Live games, it starts from the skill level you picked at signup. Daily chess does not inherit a Live rating. Because confidence is low at the start, the first games in a new time class can move the number a lot. It settles as you play more.

What Lichess says about Glicko-2

Lichess calculates ratings with the Glicko-2 method developed by Mark Glickman. Its FAQ says a new rating starts at 1500 ± 1000, and describes that as the system being 95% sure your rating is somewhere between 500 and 2500. When you are just starting out, the rating can change by several hundred points at a time. After some games against established players, the interval narrows, so each game moves you less. As time passes, Lichess says the interval increases, so later games can move the rating more quickly again.

A question mark next to a Lichess rating means it is provisional. Lichess says that happens when the Glicko-2 deviation is greater than 110: not enough rated games against similar opponents yet, or not enough recent games. Lichess says that, depending on how many games you have played, about a year of inactivity can make a rating provisional again. That "about a year" is Lichess's wording, not a stopwatch.

Those Lichess figures are Lichess's. They are not Chess.com's starting rating, and they are not the defaults in Glickman's paper.

What Glickman's paper says a rating means

Mark Glickman's Glicko-2 paper (22 March 2022) gives every player three numbers: a rating, a rating deviation (RD), and a volatility. Volatility is high when results are erratic, and low when results are steady. One way he summarizes strength is a 95% interval: the rating minus twice the RD, up to the rating plus twice the RD. His example is a rating of 1850 and an RD of 50, which is the interval 1750 to 1950. He would then say he is 95% confident the player's actual strength is between 1750 and 1950. The volatility number is not part of that interval.

Read that as a range, not as a single hidden "true rating" you are supposed to sit on. A few games inside the range are not a new level. The paper's formulas, its suggested system constant, and its unrated starting values are for someone implementing the system. They are not settings this page claims Chess.com or Lichess publish.

A coin-flip picture, not a chess model

How often is a long run "ordinary"? There is no published chess count on this page for that. Here is a picture you can check with arithmetic, labelled as a picture.

Suppose each game were an independent coin flip with a 50% chance of a loss. That is the expected score against an equal opponent if you ignore draws. One specific stretch of five games would then be five losses 1 time in 32, because 1/2 × 1/2 × 1/2 × 1/2 × 1/2 = 1/32. One specific stretch of ten games would be ten losses 1 time in 1,024. Over a few hundred games, short runs are unsurprising under that picture.

Real chess is not that coin flip. Draws exist. The next opponent's rating moves after each result. And if you are tilted, the later games are not independent of the earlier ones. The fractions only show that a streak can be ordinary variance. They are not a prediction of your next session.

This section is Live Chess on Chess.com, from its help article dated 4 April 2025, read on 1 November 2026. It is not a description of Lichess pairing, and it is not a description of Daily chess.

When you start searching, Chess.com looks for opponents within ±25 rating points. If nobody is there, it widens the range, first to ±50, and so on, so you are not waiting forever. It also tries not to pair people who are very far apart, because those games are lopsided.

"By default" matters. The ±200 line is the default limit in those rating bands, not a claim about every custom challenge. In a custom challenge you can set your own minimum and maximum opponent rating. A narrow custom range, such as ±25, is a closer match and can mean a longer wait. Unrated games use the same matching idea but are less common, so they can take longer. Color preference is considered in unrated games when you set one.

Myth: it matches your playing style, or sends someone to beat you

Playing style is not a factor in Chess.com's Live matchmaking article. The article's factors are rating, time control, how close the ratings are, widening the search, and fair-play pools. It also says, in the myth table, that being paired against the same opening is coincidence: matchmaking does not consider openings.

A winning run and a losing run can be the same length and not feel the same. The extra length usually comes from what you do next, not from a second matchmaking rule.

Long sittings add fatigue. The tenth game is not the same as the first, even when the rating band is similar. If losses bunch up late in the sitting, the rating system did not have to "find ten bad players and then ten good ones." You may have gotten sloppier: faster moves, missed tactics, a clock you stopped watching.

Tilt does the same thing on purpose, from your side of the board. The game you start right after a loss is easy to play as a replay of the last one. How to stop tilting in chess is the routine for that: a stop rule chosen while you are calm, a short review before you queue again, and leaving the board when the rule says stop. This page does not repeat that routine.

None of that is a lab result about memory. It is the practical split: some of the run is the rating band moving, and some of it is you still clicking play while tired or angry.

This is a limit you set, not a method that promises rating points. The point is to stop feeding a bad sitting.

1. Before you queue, write the stop rule. A common one, also on the tilt guide, is to stop after two losses in a row. You will not want to follow it once you are angry, which is why it is written first.
2. Also pick how many games this sitting is, and stop at that number even if you are winning. A win streak is still a reason the next opponents are stronger, and fatigue shows up on winning days too.
3. If you warm up, keep it short: a few easy puzzles or one unrated game. Should you warm up before playing chess is the longer note. A long, hard warm-up can spend the energy you wanted for the games.
4. After a loss, mark one moment in that game before the next one. How to analyze your own chess games is the short review: your own note first, the engine after.
5. When the rule is met, close the window. Coming back later is a new sitting, with the same rule.

One bad day is not a slump. This check is a way to look at your own games. The cutoffs are a simple frame, not a researched boundary. Do it in one time control, because bullet, blitz, rapid, and daily are different pools.

1. Open the last 50 to 100 games in that time control. Write the highest rating and the lowest rating in that stretch. A single evening inside that range is a swing. A range that keeps stepping down, week after week, is the longer question.
2. Mark where the losses sat in the sitting. Compare your first three games with the games after game ten. If the late games are the losses, fatigue or tilt is the first place to look, using the stop rule above.
3. On the losing run, write the opponents' ratings next to yours. If they were higher, the rating band had moved up after your wins, which is what matchmaking is described as doing. If they were lower and you still lost, the result is not "the system found stronger players."
4. Give each loss one tag you can say in a sentence: hung piece, missed tactic, time trouble, or a bad position from the opening. The tag that dominates is what to review. The counting method is how to analyze your own chess games.

What "good" means inside one pool is a separate question: what is a good chess rating. This page does not set a number of points that counts as a slump.

Worry less about a run of games this week. Look again if the rating in that time control is lower across several weeks, and especially if the same slide shows up in another time control. That is the page why am I getting worse at chess: long sessions, games started soon after a loss, a shift into faster time controls, or a new opening that has not settled.

Rating is also a noisy number. How to know if you are improving at chess is the other side: what else to count besides one week's rating. A downswing of many games in a few days can happen when you play a lot while tired. It is not, by itself, proof you lost the skill. This page has no published figure for a typical one-day drop, including a drop of about 100 points. Judge the range over a month or more in that time control.