Devig Prediction Markets: Kalshi & Polymarket | Claw Arbs
← Blog Apr 21, 2026

Devigging Prediction Markets: Fair Odds from Sharp Books

Strip the bookmaker margin. Compare to Kalshi or Polymarket. Find the edge.

Devigging is the process of stripping the bookmaker's profit margin from posted odds to estimate the implied true probability of an event. For prediction market bettors, devigging sharp sportsbook lines (Pinnacle, PS3838) gives you the best retail-accessible fair-value reference for sports events. Comparing devigged probability to the Kalshi or Polymarket contract price surfaces +EV opportunities.

This guide covers three devigging methods (proportional, Shin, logarithmic), when each one is accurate, why Pinnacle is the right reference book, how to convert devigged probability to a Kalshi cent price with fee adjustment, and how to apply the same logic on Polymarket with gas rather than percentage fees. Whether you're running value betting, arbitrage, or both, correct devigging is the foundation of the math.

What Devigging Is and Why Prediction Market Bettors Need It

Bookmakers make money by offering odds at worse-than-fair probabilities. The sum of implied probabilities across all outcomes in a market always exceeds 100%. This excess is called the overround, vig, or margin. In a two-way sports market at Pinnacle the overround is typically 2 to 3%. At soft sportsbooks it's 8 to 10%. At three-way markets the vig distributes across three prices but follows the same logic.

Definition: Overround (or vig) is the amount by which bookmaker implied probabilities exceed 100%. If Team A is listed at 1.91 decimal (52.4% implied) and Team B at 1.99 (50.3% implied), the overround is 102.7% and the vig is 2.7%. To get the bookmaker's estimate of true probabilities, you strip out the 2.7%.

Prediction market prices on Kalshi and Polymarket reflect traders' estimates of probability, not a bookmaker's margin. A 0.52 Kalshi YES price is directly interpreted as a 52% probability estimate. To value bet Kalshi against Pinnacle, you need both prices in the same format. Pinnacle's 1.91/1.99 quote has to be converted to no-vig probabilities (51.0%/49.0%) before you can compare to Kalshi's 0.52/0.48.

This conversion is what devigging is. Without it, you can't run prediction market value betting or cross-venue arbitrage reliably. Naive comparison of Pinnacle's raw implied probability (52.4%) to Kalshi's 52% would show zero edge when in fact there might be a 1.0% gap after properly accounting for the Pinnacle vig.

Devigging fits inside the broader framework of prediction market value betting. If you haven't read our pillar post on +EV strategy, start there for context on why this matters.

The Proportional Method

The proportional method is the simplest devigging approach. It assumes the vig is distributed uniformly across outcomes in proportion to their implied probability. Divide each implied probability by the sum of all implied probabilities.

Proportional formula (two-way market):
p_true_A = (1 / odds_A) / ((1 / odds_A) + (1 / odds_B))
p_true_B = (1 / odds_B) / ((1 / odds_A) + (1 / odds_B))
These always sum to 1.

Worked Example

Pinnacle NBA: Celtics 1.75 decimal, Heat 2.20 decimal.

These fair probabilities sum to 100% and represent Pinnacle's estimate of the true probability, stripped of the margin.

When Proportional Is Accurate

Proportional devigging is accurate when both outcomes are priced close to 50-50. For moneylines between roughly 1.60 and 2.50 decimal on both sides, the error is typically under 0.5% of probability. This covers the majority of sports markets.

Where Proportional Fails

For heavy favorites (odds below 1.40) and heavy underdogs (odds above 3.50), proportional devigging systematically overweights the favorite. This is because bookmakers concentrate their vig on the favorite side (where casual bettors tend to concentrate), so stripping the vig uniformly underestimates how much of the listed probability was margin on the favorite side. Shin's method fixes this.

Shin's Method

Shin (1992) published a paper arguing that bookmaker margins exist not just as profit but as protection against insider trading. Bookmakers know some bettors have better information, and they widen their implied probabilities to protect against being arbed. This insider-trading adjustment is larger for favorites (where insider advantage is more valuable) than for underdogs.

Shin's model solves a small quadratic to find the insider-trading parameter z, then uses z to estimate fair probabilities.

Shin's formula (two-way, simplified): For implied probs π_A and π_B summing to π_total, solve z = ((π_total - 1) × (π_A^2 - π_B^2)) / (π_total × (π_A^2 + π_B^2 - π_total)). Then p_true_A = (sqrt(z^2 + 4(1-z) × π_A^2 / π_total) - z) / (2(1-z)) and similarly for p_true_B.

The math is uglier than proportional but the practical effect is that Shin shades probabilities slightly toward 50-50 compared to proportional, which is empirically more accurate for extreme lines.

Worked Example (Heavy Favorite)

Pinnacle NFL: Chiefs 1.30 decimal (heavy favorite), Bengals 3.80 (heavy underdog).

Shin gives the favorite about 0.66% less probability and the underdog about 0.66% more, compared to proportional. For an arbitrage or value trade, that 0.66% can flip an edge from positive to negative or vice versa.

The Logarithmic Method

The logarithmic method (sometimes called logarithmic regression or the odds-ratio method) uses logarithms of odds to compute fair probabilities. The intuition: the vig scales multiplicatively on the odds ratio, so log-linear scaling strips it correctly for extreme lines.

Logarithmic method: For each outcome, compute log(1 / odds). Solve for a scaling factor k such that the exponentiated weighted log-probs sum to 1. The result is each outcome's fair probability. For extreme lines this handles the tail behavior better than proportional or Shin.

In practice the logarithmic method produces results close to Shin for most lines but diverges in the tails (odds below 1.10 or above 10.0). It's the right choice when you're devigging highly skewed markets, like a 90-cent political favorite or an NBA game with a 20-point spread favorite at -2000 moneyline.

The logarithmic method doesn't have a clean closed-form solution for most cases, so it's computed iteratively or via a lookup table. Production bots use pre-computed weights for speed.

Comparison Table: 3 Methods on 4 Example Lines

Here's how the three methods diverge on typical lines you'd actually encounter. "Fair prob of side A" rows show each method's estimate.

Line Odds A / B Proportional Shin Logarithmic
Coin flip 1.91 / 1.99 51.03% 51.03% 51.03%
Small favorite 1.75 / 2.20 55.69% 55.62% 55.65%
Heavy favorite 1.30 / 3.80 74.51% 73.85% 74.12%
Extreme favorite 1.08 / 11.00 91.06% 90.15% 90.45%

Takeaways: near 50-50 all three methods agree to within 0.1%. Proportional overweights favorites as lines get extreme. Shin shades probabilities slightly toward 50-50 for extreme favorites, which empirical studies support as more accurate. Logarithmic splits the difference and performs best in the extreme tail.

For a typical Kalshi or Polymarket value betting bot, the pragmatic choice is Shin as the default with proportional as a fallback when Shin fails to converge. Logarithmic is used selectively for extreme markets.

Why Pinnacle and PS3838 Are the Right Reference Books

Devigging is only as accurate as the book you're devigging from. Pinnacle and PS3838 (which uses Pinnacle's pricing engine) are the correct references for three reasons.

Tight Margins

Pinnacle runs on 2 to 3% margin on mainlines, compared to 8 to 10% at soft books like DraftKings or FanDuel. Smaller margins mean the vig adjustment you apply is smaller, so any devigging error is proportionally smaller. A 0.3% devigging error is a much bigger problem on a soft book line (where your resulting fair estimate could be 2-3% off) than on a Pinnacle line (where the same 0.3% error leaves you within 0.5% of fair).

Sharp Action Welcomed

Pinnacle accepts sharp money without limiting. This means Pinnacle's prices incorporate information that soft books' prices don't, because soft books reject the bets that would otherwise move their lines. If professional bettors know something, Pinnacle's line knows it. Soft books don't.

Fast Line Adjustments

Pinnacle moves lines within seconds of new information (injuries, lineup changes, weather). Soft books can lag by minutes or hours. When you use Pinnacle as reference, you're devigging the most up-to-date sharp consensus. When you use a soft book, you might be devigging a line that's already stale.

For a broader view of which sportsbooks support which strategies, see our best sportsbooks for arbitrage betting guide. Pinnacle is essential not just for arbitrage but for any strategy that requires a clean sharp reference.

Beyond Pinnacle: Multi-Book Sharp Aggregators

Pinnacle alone is a strong reference. A multi-book sharp odds aggregator is stronger. Aggregators blend Pinnacle with 10 other sharp books (Pin88, BetIsn, BetDEX, Punter.io, and others) into a single tighter true-odds line. Three concrete advantages over single-book devigging: the aggregate is closer to true probability than any individual book (independent sharps disagree slightly, blending averages away the noise); the aggregate keeps streaming when one book pulls a market live in-play (you don't lose your reference during a goal); and the aggregate emits sharp-line freeze signals within milliseconds when any constituent sharp suspends a market, 5 to 15 seconds before soft books react. For a bot devigging at scale, the upgrade in edge quality and freeze-signal speed is meaningful. Claw Arbs ships an 11-book aggregator built in and uses it as the default value-bet reference, with single-book Pinnacle / PS3838 as the per-book fallback.

Applying Devigged Odds to Kalshi

Once you have a devigged probability from Pinnacle, compare it to the Kalshi contract price to find value. The workflow:

  1. Devig Pinnacle (get fair probability for the outcome that matches the Kalshi contract)
  2. Convert probability to Kalshi cents (simple: a 0.55 probability corresponds to 55 cent contract price, since Kalshi contracts are priced in cents equivalent to probability)
  3. Compute the Kalshi fee: fee = tier × P × (1 - P) where P is the current Kalshi price in dollars and tier is your volume-based fee rate (3 to 7%)
  4. Compute net edge: net_edge = (fair_price - kalshi_price) - fee, or in relative terms, net_edge_relative = net_edge / kalshi_price
  5. Compare to your minimum edge threshold (typically 2% net for Kalshi)

Example

Pinnacle Chiefs 1.75, Heat 2.20 (from above). Shin devig gives Chiefs fair = 55.62%. Kalshi "Chiefs win" YES at 52 cents.

Fair cents: 55.62.

Kalshi price: 52.

Gross edge: 3.62 cents.

Fee at 7% tier on 52-cent contract: 0.07 × 0.52 × 0.48 = 1.75 cents.

Net edge: 3.62 - 1.75 = 1.87 cents, or 3.60% relative to 52 cent price.

3.60% net edge is above threshold. +EV play. This is exactly the workflow our Kalshi value betting guide walks through for each market category. Use the Kalshi fair odds calculator to automate the full computation.

Applying Devigged Odds to Polymarket

For Polymarket the workflow is similar but fee math changes. No percentage fee on wins, just Polygon gas. Gas cost depends on your stake size.

  1. Devig Pinnacle (get fair probability)
  2. Convert to Polymarket decimal (0.5562 probability = $0.5562 YES price equivalent)
  3. Gross edge = fair price minus Polymarket price
  4. Subtract gas cost as a percent of stake (typically negligible above $200 stakes)
  5. Net edge vs. your threshold (typically 1.5% relative for Polymarket)

Example

Same Pinnacle line. Polymarket Chiefs YES at $0.53. Shin devig fair: 0.5562.

Gross edge: $0.5562 - $0.53 = $0.0262, or 4.94% relative.

Gas at $500 stake: $0.10, or 0.02% of stake.

Net edge: 4.94% - 0.02% = 4.92%.

Strong +EV play. Our Polymarket value betting guide walks through similar workflows for politics and macro markets where the reference isn't a sportsbook.

Live Example: Full Devig + Kalshi Comparison + Edge Decision

Let's run through an end-to-end example combining everything above. Setup: Pinnacle Premier League, Arsenal 2.60 decimal, Draw 3.20, Manchester City 2.70 (three-way market). Kalshi sports contract "Arsenal wins match" YES at 34 cents.

Step 1: Implied Probabilities from Pinnacle

Step 2: Apply Devig Method

For three-way markets the choice matters more than for two-way. Proportional redistributes vig uniformly:

A multiway Shin method (extension of the two-way version) would give slightly different numbers:

Use the conservative estimate: Arsenal fair probability is ~35.9 to 36.0%.

Step 3: Compare to Kalshi

Kalshi YES at 34 cents. Fair value: 35.9 cents.

Gross edge: 1.9 cents.

Fee at 7% on 34-cent contract: 0.07 × 0.34 × 0.66 = 1.57 cents.

Net edge: 1.9 - 1.57 = 0.33 cents, or 0.97% relative.

Step 4: Decision

0.97% net edge is below the 2% Kalshi threshold. Skip this trade. If the Kalshi price dropped to 32 cents (unchanged Pinnacle), re-run: gross edge 3.9, fee at 32 cents is 1.52, net 2.38 cents or 7.4% relative. Now it's a clear +EV trade.

This example illustrates why devigging accuracy matters. If you used proportional and got 36.03% fair, vs Shin and got 35.90%, the difference is 0.13%. On most trades that's irrelevant. On marginal trades near your threshold, it flips the decision. Production bots that care about consistent edge should use Shin's method or better.

Automating Devigging

Manual devigging is fine for verifying trades, but running a live value betting strategy requires automation. A devigging engine in a value betting bot:

The Kalshi fair odds calculator lets you test Shin vs proportional outputs against hypothetical lines. The prediction market EV calculator helps you verify the downstream edge and expected value math. If you want the whole pipeline running in production with tested devig methods, sharp reference polling, and automated execution, Claw Arbs includes all of this out of the box.

For additional context on how devigging fits into broader prediction market strategies, visit the prediction market hub.

Frequently Asked Questions

What's the difference between proportional and Shin devigging?

Proportional devigging divides each implied probability by the sum of all implied probabilities, distributing the vig uniformly across outcomes. Shin devigging (from Shin 1992) assumes the vig exists because the bookmaker protects against insider trading, so favorites carry more vig relative to underdogs. Proportional is accurate near 50-50 lines, Shin is more accurate for heavy favorites (odds below 1.40 or above 3.50). For most sports markets they produce similar results within 0.5 percent.

Which devigging method is most accurate?

No single method is universally most accurate. Proportional works well for balanced two-way markets (both odds between 1.60 and 2.50). Shin is more accurate for heavy favorites and heavy underdogs. The logarithmic method handles extreme lines best but is less intuitive. Professional traders often use Shin as the default and fall back to proportional for markets with small vig. For sub-1% accuracy you ultimately need to weight multiple methods or calibrate against historical outcomes.

Why is Pinnacle the preferred reference for devigging?

Pinnacle (and PS3838, which uses Pinnacle's pricing engine) runs on 2 to 3 percent margin versus 8 to 10 percent at soft sportsbooks. They accept sharp money without limiting and adjust lines within seconds of new information. This means Pinnacle's quotes are the closest retail-accessible estimate of true probability. Soft books have larger margins, reject sharp action, and often lag sharp consensus by minutes, which makes their lines much less useful as a devigging reference.

Can I devig prediction market odds to find value on sportsbooks?

In theory yes, but in practice it rarely works. Prediction markets like Kalshi and Polymarket tend to lag sharp sportsbooks, not lead them, so using prediction market prices to find value on sportsbooks gives you stale information. The direction that works is the opposite: devig sharp books to find value on prediction markets. The exception is major political markets during US hours where Polymarket may lead sportsbook political offerings, but this is a narrow use case.

What's the minimum edge needed after devigging?

For Kalshi value betting, a 2% net edge after fees (which requires roughly 3.5 to 4% gross edge on mid-priced contracts near 50 cents) is a reasonable floor. For Polymarket, 1.5% net is usable because there are no percentage fees. For arbitrage plays that cross venues, edges as small as 0.5% after costs are tradeable because you lock in the spread rather than taking directional variance. Below these thresholds variance dominates and expected value is unreliable.

The Bottom Line

Devigging prediction markets is the technical foundation for every value betting and cross-venue arbitrage strategy that uses sharp sportsbooks as reference. Without accurate devigging, your fair-value estimates will be systematically wrong, your edge math will lie to you, and your bot will fire on non-edges all day. With the right method (Shin as default, proportional for balanced lines, logarithmic for extremes) you get fair probabilities accurate within 0.5% for most sports markets.

Combine devigged Pinnacle with fee-aware Kalshi math and you have a systematic +EV process. Add Polymarket with gas-aware math and you double your venue coverage. Automate it all and you scale. The sibling guides cover each piece: Kalshi value betting for Kalshi specifics, Polymarket value betting for Polymarket specifics, and the prediction market value betting pillar for the overall strategy.

When you're ready to put this in motion, download Claw Arbs. The devigging engine, sharp reference polling, fee math, Kelly sizing, and execution pipeline are all built in. You configure the strategy parameters; the bot handles the arithmetic in real time.

Ready to find arbitrage opportunities?

Claw Arbs detects and executes arbs across sportsbooks and prediction markets in ~20ms.

Get Claw Arbs