#Where they are
The gear in the sidebar opens Settings, in four tabs: Play, Analysis, Board and Engine. Everything is stored in this browser, so it follows the device rather than the account.
The language is picked from the globe beside the dialog's title, above the tabs, because it belongs to none of them.
A few settings are not in that dialog at all, because they belong beside the work: the Tutor's coaching level is set from the gear inside the analysis pane, the Study and Transcribe settings from their own surfaces, and the Replay ones from the transport bar under the board. They are listed here with the rest.
Every table on this page is generated from the app's own settings catalogue, which is also where the defaults in the code come from. A default that changes changes here too, in the same commit, with nobody having to remember this page exists.
#Play
The game in front of you.
| Setting | What it does | Default |
|---|---|---|
| Play depth | How far the engine looks ahead when it plays and when it gives you a hint. Deeper is stronger and slower. Choices: 2-ply, 3-ply | 3-ply |
| Jacoby rule | Money games only. Under the Jacoby rule a gammon scores as a plain win until someone turns the cube, which makes doubling worth more and playing on for a gammon worth less. Match play never uses it. Auto follows whatever rule the match itself recorded, which is what an imported session was actually played under. Choices: Auto, On, Off | Auto |
| After a game | Auto-continue starts the next game a few seconds after the recap card appears, unless you touch anything. It waits for the game review to finish first. Choices: Wait for me, Auto-continue | Wait for me |
| Review each game | Analyse each finished game on your device while the recap card is up, so the session is already analysed when you open it. In unlimited sessions this is what keeps a long session from costing one long pass at the end. With the Tutor off the numbers stay hidden behind Show review. Choices: Unlimited only, Always, Never | Unlimited only |
| Player name | The name recorded as yours in match files and exports. An account supplies one, so the field is only offered while you are signed out. | empty |
| Coaching level | How much the Tutor says while you play, set from the gear inside the analysis pane. Choices: Hands-off, Hints, Tutor | Hints |
| Hold from | Which mistakes the Tutor stops your hand on. Only meaningful at the Tutor level, where it can hold at all. Choices: From errors, Blunders only | From errors |
| Keep the analysis open | Under Tutor, leave the read-out up across every decision instead of opening it each turn. Nothing is drawn on the board until you pick a move from the list. Choices: On, Off | Off |
| Follow the analysis | Show the analysis of the last move played, yours or your opponent's, beside the board as the game goes on. After each of your moves the opponent waits, with both clocks stopped, until you press Continue or Space. Your cube decisions are followed too: a roll with the cube yours to turn, a double, a take and a pass. Pick a move from the analysis to look back at it. Choices: On, Off | Off |
| Clock | The match clock, as reserve per point, delay and increment in seconds. Set when you start a match; empty means no clock. | no clock |
#Analysis
The numbers put on a game afterwards. Depth is the setting that matters most: see the note on ply numbers at the end.
| Setting | What it does | Default |
|---|---|---|
| Match analysis depth | The depth a whole-match pass runs at on this device. A match is hundreds of decisions, so a rung here costs hundreds of times what it costs on one position. Choices: 2-ply, 3-ply | 2-ply |
| Position analysis depth | The depth a single position gets, in Study and wherever one position is asked about. Deeper than the match default, a single position being cheap. Choices: 2-ply, 3-ply | 3-ply |
| Analysis threads | How much of your device on-board analysis may use. Higher settings analyse a match faster by spreading it over more of your CPUs, and use more memory and battery. Play is never slowed down: it keeps its own engine. Choices: Low, Medium, High | Medium |
| Equity display | How much precision the equity numbers carry. Decimal modes show points; Cents shows points x100 to one decimal place, a cent being a hundredth of a point. Win probabilities and error rate are unaffected. Choices: 0.0001, 0.001, 0.01, 10.0 | 0.001 |
| Move list | Detailed prints each candidate move's chances under its notation: W is the chance of winning, and G and B are gammons and backgammons as win/loss pairs. Backgammons are left out where none is possible. Choices: Compact, Detailed | Compact |
| Match unit | Which unit a match analysis is shown in. MWC is match-winning chance, the currency the analysis is made in; EQ is the same answer normalized into points. Error rates and the blunder labels are in normalized equity either way. Money games have no MWC. Choices: EQ, MWC | MWC |
| Review scope | Whose mistakes the review walks when you step from error to error. The start of a game is always a stop, being context rather than anyone's mistake. Choices: Both players, Mine, My opponent's | Both players |
| Where analysis runs | Whether a match is analysed in this tab or on our servers. The answer decides which engines are offered, which depths are reachable and which plan applies, so the run modal opens on whichever you chose last. Choices: This device, Our servers | This device |
#Board
How the whole thing looks and sounds.
Two of these are about how a decision that has already been played is drawn, and they are set from the menu beside the board rather than from this dialog. Move display chooses between the position before the play, the play itself as arrows over that position, and the position it leads to.
Showing the best move too puts both moves on the board at once, in two colours. The best move is green, in every theme. The move that was played takes the colour of its own verdict, the same ladder the move list glows by: red for a blunder, orange for an error, amber for an inaccuracy, and a plain grey where it barely cost anything, because a play two thousandths off best is not a mistake and should not be painted as one. Only where the two moves differ, so an overlay is never a move drawn on top of itself.
| Setting | What it does | Default |
|---|---|---|
| Theme | The colour palette across the whole app. Choices: Slate, Burrow, Warren, Bristle, Hedgerow | Warren |
| Animation speed | How fast a checker slides to where it was played. Off moves it there at once. Choices: Off, Fast, Normal, Slow | Fast |
| Pip count | Whether each player's pip count is shown on the board. On hover shows a faint "..." and reveals the count when you hover over it. Choices: Show, On hover, Hide | Show |
| Point numbers | Whether the point numbers (1-24) are shown along the edges of the board. Choices: Show, Hide | Show |
| Sound | Board audio while you play: checkers, dice and the cube. Silent turns it off. Choices: Wood, Silent | Wood |
| Volume | How loud the board is, from silent to full. | half |
| Move display | What the board shows for a decision that has been played: the position before it with nothing over it, that same position with the move drawn over it, or the position the move leads to. Choices: Before the play, Move arrows, After the play | Move arrows |
| Show the best move too | On an analysed decision, draw the best move over the board as well as the move that was played. The best move is green; the move that was played takes the colour of its own verdict - red for a blunder, orange for an error, amber for an inaccuracy, and a plain grey when it barely cost anything. Only where the two moves differ, and only under Move arrows above. Choices: On, Off | Off |
| Flip vertically | Swap who sits at the top of the board and who sits at the bottom. Choices: On, Off | Off |
| Flip horizontally | Mirror the board left to right, which moves the home boards and the tray to the other side. Choices: On, Off | Off |
| Swap checker colours | Draw each player in the other physical colour. Nothing about the game changes, only which set of checkers is which. Choices: On, Off | Off |
#Language
The globe at the top of the Settings dialog, beside its title, picks the language of the interface. It sits outside the four tabs so that someone who landed in a language they cannot read can still find the way back without guessing which tab it is filed under, and every language in the list is written in itself. The globe only appears once there is more than one language to choose from.
| Setting | What it does | Default |
|---|---|---|
| Language | The language of the menus, buttons and labels. Help pages, legal documents and move explanations stay in English. Choices: Default, English, Deutsch, فارسی, Français, עברית, Nederlands, Русский, Türkçe | Default |
#Study
Set from the Study surface itself rather than from the Settings dialog.
| Setting | What it does | Default |
|---|---|---|
| Tree style | How a study's variations are drawn: one row per node, as a branching game score, or as that score flattened into paragraphs. Choices: List, Game score, Flattened score | List |
| Move notation | Whether moves are written out in full or compressed. Choices: Compact, Normal | Normal |
| Equity colouring | Tint each row of the tree by how its equity compares with a reference node you pin. Off by default: a tree tinted green and red before you have asked for it reads as severity. Choices: On, Off | Off |
#Transcribe
Set from the transcriber, and scoped to it. The text-size ladder in particular is there because that is the surface whose labels are dense: a rail of dice, scores and notation read at arm's length from a board.
| Setting | What it does | Default |
|---|---|---|
| Text size | How large the transcriber's own labels are drawn. The rail of dice, scores and notation is read at arm's length from a board, often on a phone, so it has a size ladder of its own. Choices: Small, Medium, Large, Extra large | Medium |
| Video pre-roll | In video mode, how far before a stamped move the player seeks to. A transcriber marks a play a beat after seeing it, so a pre-roll is what lands the seek on the roll rather than on what follows it. | 3 seconds |
#Replay
A match open in the analysis view can advance itself instead of waiting for the
arrow keys. The bar is off until you ask for it, from the same menu beside
the board that the two rows above are set from: most people read a match with
the arrow keys, and a control most people never press does not belong under
every board. Turn it on and it stays on, and the keys it claims - space to play
or pause, [ and ] for the speed, M to mute - are claimed only while it is
showing. Choosing Replay this match from that menu also starts it playing,
on a phone as well as a desktop: on a phone the bar sits directly under the
board, or at the top of the analysis column when the phone is on its side. The
bar carries play, the position, the speed, the voice and mute; the pacing, the
stops and the voice's details are behind its options button. The runtime
estimate is printed on the bar.
The checkers move rather than teleporting, one at a time, at whatever animation speed is set under Board above. Where part of a play was not a choice - entering from the bar, a play every ranked alternative also makes - that part goes down straight away and the rest follows after the pause, which is the shape of somebody actually playing it. The runtime printed on the bar counts the thinking and not the flying, so it is a floor rather than a stopwatch.
Recorded pacing gives a move more screen time when the player spent more time on
it, compressed so that a long tank reads as a pause rather than a stall. It
needs a match with a recorded clock, which means one played or transcribed here;
an imported .mat, .xg, .bgf or .sgf carries no clock and replays on the
even beat whichever pacing is chosen.
It can also say what happened out loud, using the speech synthesiser your browser already has. The speech-bubble button on the bar turns the voice on and off, and says something straight away so you know it works. It reads in an English voice whatever language your device is set to, and you can pick another English voice in the options. While a replay plays, the screen is kept awake; switch to another app or tab and the replay pauses where it was. Nothing is sent anywhere for that: the voice is generated on your own device, and which voices exist depends on the device rather than on us. The same words appear on the bar whether the voice is on or off.
A replay stops where you asked to be stopped: on a move at or below the quality floor, on a cube action, and on each game's turning point if you want it. Whose mistakes count is the review seat you already set for the next-error buttons, in Analysis above, so the two can never disagree. At a stop the move that was played goes on the board, and a beat later the move that should have been is added beside it, in green: both at once, which is what a reader stopped at a blunder came for.
At a stop it can also ask for a sentence explaining the position, the same one the Why? button gives you. That is the one setting here that can cost you something: on the free plan explanations are limited per day and each new move costs one, so it is off until you turn it on. A move already explained today is free to hear again, which is what makes re-watching a match free.
| Setting | What it does | Default |
|---|---|---|
| Replay bar | Show the transport bar under the analysis board, which plays the match through by itself. Off by default: the arrow keys are how most people read a match, and the bar is turned on from the menu beside the board when it is wanted. Choices: On, Off | Off |
| Replay pacing | How long each move stays on screen while a match replays itself. Recorded gives a move more time when the player spent more time on it, compressed so a long tank reads as a pause rather than a stall. Even gives every move the same beat, and is what a match with no recorded clock gets whichever you pick. As played is the literal think time. Choices: Recorded, Even, As played | Recorded |
| Replay speed | How fast a self-driving replay runs. The estimate printed beside the transport is the runtime at 1x, and this divides it exactly. Choices: 0.5x, 0.75x, 1x, 1.5x, 2x, 3x | 1x |
| Stop on | How bad a move has to be for the replay to halt on it. The same ladder the move list glows by, and a move that failed to double counts at whatever that omission cost. Choices: Never stop, Inaccuracies and worse, Errors and worse, Blunders and worse, Whoppers only | Blunders and worse |
| Stop on cube actions | Halt on every double, take and pass, however well played. A cube turn changes what the rest of the game is worth, which is worth seeing whether or not it was a mistake. Choices: On, Off | On |
| Stop on each game's turning point | Halt on the costliest decision YOU made in each game, even when it is under the floor above. Your own moves only, so a review scoped to your opponent's mistakes never uses it. Choices: On, Off | Off |
| How long to stop for | What a stop does: wait until you press play again, or pause for a set time and carry on. Waiting is right when you are studying; a timed pause is right when you are watching. Choices: Until I continue, 3 seconds, 6 seconds, 10 seconds | Until I continue |
| Show the better move | At a stop, show the move that was played, then add the best move beside it a beat later, in green, so the two can be compared without touching anything. Choices: On, Off | On |
| Say what happened | Read each move out loud while the replay runs, using your own browser's voice. Also the speech-bubble button on the bar. Nothing is sent anywhere: the speech is generated on this device. The same words are shown on the bar whether it is on or off. Choices: On, Off | Off |
| How much to say | Brief says the move and its verdict. Normal adds the score at the start of each game. Full adds what the error cost, in thousandths of a point. Choices: Brief, Normal, Full | Normal |
| Speaking rate | How fast the voice reads. Separate from the replay speed, which moves the board: a fast reader can still want a slow replay. Choices: Slow, Relaxed, Normal, Quick, Fast | Normal |
| Explain a stop | Ask for a sentence explaining the position when the replay stops on a move this bad or worse, and read it with the rest. On the free plan explanations are limited per day and each new move costs one, so this is off until you turn it on. A move already explained today is free to hear again. Choices: Never, Blunders and worse, Errors and worse, Inaccuracies and worse, Every stop | Never |
#Engine
The local model list is the neural network that runs in your browser to play, hint and analyse. Stronger models give more accurate moves and equities, and cost more memory and time. The default is loaded for you; you can switch to another published model, or import one you have trained yourself.
Which models you may choose from depends on your plan, because some are published for download and some are not. See Accounts and plans.
The model is not in the table above because it is not a preference with a default: it is whichever engine you last chose, remembered so the next launch works offline.
#A note on ply numbers
HedgeHog counts plies the way eXtreme Gammon and BGBlitz do, where 1-ply means a direct evaluation with no search and each further number adds a level of look-ahead. GNU Backgammon counts from zero, so its "2-ply" is our 3-ply. The number shown everywhere in this app is the display one.