Manual rating adjustments

Ratings build themselves from recorded scores — that's almost always the right call. Two situations are the exception: seeding a brand-new player so they don't start from a flat default, and correcting a rating that's gone clearly wrong. Here's how each works, and what a hand-set value does once real games start landing.

Seeding a new player

When you add a player with a known level, type a starting rating into the rating field on their player record — on your group's scale. The mixer treats that as their starting point and adjusts it from there.

Decimals work with either separator: type 4.2 or 4,2 and both read as the same number, so a comma-decimal keyboard (French and other locales) no longer turns 2,5 into 25 behind your back.

The scale part matters. If your group runs on the standard 1,500 scale, a DUPR-style number like 3.5 is not a rating there — saved as one, it would rank that player hundreds of times weaker than everyone else and skew every game they play. So:

• Your group uses standard-scale ratings (the default): seed with a number near your other players (say 1,350–1,650), and keep the DUPR number in the Skill / DUPR note field instead — it shows as a badge by their name and never touches the math. If you type a skill-level number into the rating field anyway, a dialog will stop you and offer to move it to the note for you.
• You want to actually run on DUPR-style numbers: that works too — first calibrate your group's rating settings (Group Settings → Ratings: set the default to ~3.5, then apply the recommended scale and K-Factor — all three have to move together or the rating math drifts), then seed everyone in those units. Once your roster is coherently on the small scale, the warnings stay quiet.

The shortcut for a brand-new group: if a group has recorded no games yet and you seed everyone with DUPR-style numbers, PickleFriend recognizes that the whole group is running on that scale and quietly calibrates the rating settings for you — no dialog, no confirmation. Group Settings → Ratings will simply show the DUPR-matched default, scale, and K-Factor afterward. It's symmetric: if you later delete those ratings (or change them all back to standard-scale numbers), the settings revert to the standard 1,500 scale on their own. This works the same in fixed-partner (team) groups — team seeds derive from the player seeds, so calibrating on the players carries through. Two guardrails to know: it only ever kicks in when every rating is DUPR-style (a mix of skill numbers and standard ratings is treated as a typo, not a scale, and stays flagged), and it never touches a group whose rating settings you've customized yourself — your own configuration is always respected.

The warning catches typos in both directions. A number that's far too low — a skill level in a standard-scale group — ranks the player below everyone, as above. A number that's far too high — the classic being a missing decimal point, so 2.5 becomes 25 — gets its own heads-up: "that's much higher than everyone else, probably a typo." (The high-side warning just asks you to double-check the number; there's no note-move offer, because a value like 1500 isn't a skill level worth keeping.) The high check is relative to the rest of your roster, so it only fires once your group has a settled set of ratings to compare against — a brand-new group or a genuinely wide-ranging one won't get false alarms.

If a mis-typed starting rating does slip through, the dashboard flags it on your next visit and offers a per-player fix that resets the outliers to a sensible starting value and recalculates every game. For a too-low skill number, the fix doesn't discard what you typed: the original value is kept as that player's Skill / DUPR note (unless they already have one), so the number still shows as a badge by their name. A too-high ×10 typo is just corrected — nothing is kept as a note, since a stray 1500 isn't a skill level worth remembering.

Seeding matters most in rating-balanced mixing and in ladders — a sensible starting point means tight matchups from the first game rather than a few rounds of calibration. Leave someone unrated and they aren't treated as a zero; "no rating" is its own state until they've played enough to establish one. See How Elo works.

What happens once games are recorded

A hand-set rating is a starting point, not a permanent lock. Once a player has recorded games, the rating reflects those results and the system protects that earned value. Overwriting the rating directly on a player who already has games is blocked — their other profile changes save, but the rating field is left alone.

A direct overwrite would silently discard everything their games taught the system. So:

You want to…	Use…
Set a level for a player with no games yet	The rating field on the player record — it seeds cleanly
Nudge the rating of a player who has played	A manual adjustment (below), not a direct edit
Fix a wrong score	Editing the game, not the rating

Correcting an outlier with a manual adjustment

When an established rating drifts wrong — a string of flukey results, or a known data-entry mess — apply a manual adjustment: a deliberate bump up or down with a reason attached. Unlike a raw overwrite, an adjustment is recorded as its own event in the player's history, slotted in chronological order alongside their games, and it can be reversed later if you change your mind. Nothing the games taught the system gets thrown away — the adjustment sits on top of that history, and the audit trail shows exactly what you changed and why.

Adjusting a rating vs. editing a game

These solve different problems. Reach for the wrong one and you'll confuse yourself.

• Editing a game fixes the input — a score typed wrong, the wrong winner, the wrong players on a court. The rating then recomputes from the corrected result. This is almost always what you want when something specific was entered incorrectly. See Editing recorded games.
• A manual adjustment changes the rating itself, independent of any one game — for seeding gaps or correcting a rating you believe is off for reasons the scores don't capture.

Don't over-tweak

The rating system is built to self-correct: a player who's stronger than their number will keep winning and climb on their own within a few games. Constant hand-nudges fight that process and make standings harder to trust. Seed once, fix genuine outliers, then let the games do the work.

Recalculate All Ratings

Found in the Advanced section at the bottom of the Games page, "Recalculate All Ratings" is the reset valve for a group whose ratings have gotten genuinely tangled — a batch of bad data, a settings change you want applied retroactively, or a season restart where everyone keeps their history but you want the numbers re-derived cleanly.

Here's what it does:

1. Resets every player to their seed rating (or 1,500 if none was set), wiping current game counts and stats.
2. Marks every game for re-processing — in chronological order, by a background cron job that re-runs Elo as if it were seeing each game for the first time.
3. Re-queues any manual adjustments alongside those games, so adjustments you've recorded are folded back into the timeline at the right point.

The source records for manual adjustments are preserved — a recalculation replays them, it doesn't delete them. Games are replayed too, not deleted.

One condition: if there are games still being processed when you hit the button, the recalculation is blocked until the queue clears. The error message tells you how many are pending.

For large groups with many recorded games, ratings will be temporarily at their seed values while the background job runs. Players may see their numbers mid- replay before the cron finishes — this is normal.

When to use it

• You changed a group-wide setting (K-factor, scale, win bonus) and want the history re-scored under the new rules.
• A bulk of games were entered incorrectly, you've fixed them, and you want a clean pass over everything.
• You suspect drift has accumulated from a setting that's since been corrected.

Don't use it for a single bad score — edit that game instead, and let the system handle the targeted re-run.

Delete All Games

The "Delete All Games" button sits next to Recalculate in the same Advanced section. Unlike Recalculate, this permanently removes all games AND all manual adjustments, and resets ratings to seed values. It requires you to type DELETE in a confirmation prompt. Use it only if you're starting completely fresh — there's no undo.

What's next

• How Elo works — why score-based ratings self-correct, and what placement games do.
• Editing recorded games — fix a wrong score so the rating recomputes correctly.
• Placement games — the early-game period when ratings move faster to find a player's level.