> ## Documentation Index > Fetch the complete documentation index at: https://txline-docs.txodds.com/llms.txt > Use this file to discover all available pages before exploring further. # Odds Record Fields > What each field of an odds record means, including QuoteSigma, the price and probability corridors, and Confidence This page explains the fields of an odds record, as in `OddsPayload` in the [Odds API Reference](https://txline-docs.txodds.com/api-reference/odds). Fields marked optional can be missing, null, empty or `"NA"`. Treat any of these as not provided for that record, not as zero. ## Record Fields | Field | Type | Meaning | | - | - | - | | `FixtureId` | integer | The fixture the record belongs to. | | `MessageId` | string | Unique id of the odds message. | | `Ts` | integer | Record timestamp. | | `Bookmaker`, `BookmakerId` | string, integer | The source of the quote. | | `SuperOddsType` | string | The market type. | | `MarketParameters` | string, optional | Market parameters, such as a handicap or total line. | | `MarketPeriod` | string, optional | The period the market covers. | | `GameState` | string, optional | Game state when the quote was made. | | `InRunning` | boolean | `true` once the fixture is in play. | | `PriceNames` | string array | One name per selection. Every array below is in this order. | | `Prices` | integer array | Decimal odds ×1000. `1909` is 1.909. | | `PricesLow`, `PricesHigh` | integer array, optional | Lower and upper bound of each price, on the same ×1000 scale. | | `Pct` | string array, optional | Implied probability of each selection, as a percentage with three decimals, such as `"52.632"`. `"NA"` for quarter-handicap lines. | | `PctLow`, `PctHigh` | string array, optional | Lower and upper bound of each probability, in the same format as `Pct`. | | `QuoteSigma` | integer, optional | Uncertainty of the quote, as a probability standard error ×1000. See [QuoteSigma](#quotesigma). | | `Confidence` | integer array, optional | A 0–100 score per selection. See [Confidence](#confidence). | ## De-Margined, Aggregated Prices `Prices` are de-margined: the bookmaker's overround has been removed using power de-margining, so the probabilities implied by the selections sum to 1. The prices are fair odds, not the margined odds a bookmaker offers. Stable Price aggregates lines across bookmakers. A record carries one `QuoteSigma` for the whole aggregate, not one per contributing bookmaker. Two `Bookmaker` values carry Stable Price records, and both are de-margined: | `Bookmaker` | Fields | | - | - | | `TXLineStablePrice` | De-margined, and calculated with the additional fields: `QuoteSigma`, `PricesLow`, `PricesHigh`, `PctLow`, `PctHigh` and `Confidence`, as well as `Pct`. | | `TXLineStablePriceDemargined` | De-margined, with `Pct`. The additional fields are empty. | Read `QuoteSigma`, the `Low`/`High` bounds and `Confidence` from `TXLineStablePrice` records. ## QuoteSigma `QuoteSigma` is the standard error of a de-margined quote, in probability units, stored as an integer scaled by 1000. A value of `24` means a standard error of 0.024, or 2.4 percentage points. It combines two independent sources of uncertainty: * **Overround.** A bookmaker's margin is spread over the selections in a way we can't observe. The uncertainty from this is the overround per selection. * **Tick rounding.** Raw odds are rounded to three decimals, which moves the implied probability by a small, known amount. ```text theme={null} q_i = 1 / raw decimal odds of selection i overround = max(0, sum(q_i) - 1) n = number of selections (2 or 3) sigma_vig = overround / n sigma_quant = (1/n) * sum( 0.0005 / rawOdds_i^2 ) sigma_quote = sqrt(sigma_vig^2 + sigma_quant^2) QuoteSigma = max(1, round(1000 * sigma_quote)) ``` Because it is a probability and not a price, it is always between 1 and 666. `Prices` values are always above 1000, so the two can't be confused. The floor of 1 applies to a book with no margin. | Market | Typical `QuoteSigma` | | - | - | | Sharp pre-match main line (about 2.4% margin) | 12 | | Standard live two-way line (about 4.8% margin) | 24 | | Standard three-way line (6–8% margin) | 20–27 | | Wide in-running or deep alternate line (8–12% margin) | 40–60 | > [!NOTE] > > `QuoteSigma` is in probability space, so the same value means different things at different prices. A value of 20 is tight on an even-money line and very wide on a 50.00 longshot. Use `Confidence` or the `Low`/`High` bounds when you need a measure that accounts for the price. ## Price and Probability Bounds `PricesLow`/`PricesHigh` and `PctLow`/`PctHigh` give the one-sigma range in which the true fair price or probability plausibly lies. They are derived from `QuoteSigma` in log-odds space, which keeps every bound strictly between 0 and 1 for probabilities and strictly above 1.0 for decimal odds. A plain `Pct - QuoteSigma` would go negative on a longshot. ```text theme={null} c = probability of selection i, from Pct sigma_logit = sigma_quote / max(0.0001, c * (1 - c)) M = exp(sigma_logit) PricesLow = 1 + (odds - 1) / M PricesHigh = 1 + (odds - 1) * M ``` * The bounds bracket the price: `PricesLow < Prices < PricesHigh`, and `PctLow < Pct < PctHigh`. * A higher price means a lower probability. `PricesLow[i]` therefore pairs with `PctHigh[i]`, and `PricesHigh[i]` pairs with `PctLow[i]`. * The range is asymmetric away from even money. It is about symmetric near 50%, and skews towards longer prices on a longshot. For an even-money selection quoted 1.909/1.909, `QuoteSigma` is `24`, the price is 2.000, `PricesLow` is 1.909 and `PricesHigh` is 2.100. `PctLow` is 47.619, `Pct` is 50.000 and `PctHigh` is 52.383. ## Confidence `Confidence` turns the uncertainty into a single 0–100 score per selection, where higher means a tighter quote. It uses the log-odds standard error above, so it is comparable between favourites and longshots: ```text theme={null} Confidence = round( 100 / (1 + (sigma_logit / 0.20)^2) ) ``` A `sigma_logit` of 0.20 scores 50. | Selection | `QuoteSigma` | Price | `Confidence` | | - | - | - | - | | Even-money main line | 24 | 2.000 | 81 | | Longshot on a 1.015/15.000 line | 26 | 46.916 | 3 | The same `QuoteSigma` gives very different scores here. A 2.6-point error is small on an even-money line and larger than the probability itself on a longshot. ## Using the Bounds * **Valuation.** `PricesLow` to `PricesHigh` answers "where could the true fair price plausibly lie?" * **Execution.** A price slightly above `Prices` may be inside the noise. To require an edge of `k` standard deviations, the offered price must exceed `1 + (odds - 1) * exp(k * sigma_logit)`. At `k = 1` this is `PricesHigh`. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.