In LMSR and related designs, b is the dial between “toy market” and “institutional depth.” Too low and prices jump on fifty dollars; too high and the operator locks capital for bounded benefit.
This chapter explains b in trader and operator terms—with numbers—and how a deep regulated book differs from a large b on chain.
What b controls
b scales how sensitive prices are to net shares sold. Small b means each trade moves implied percent a lot (steep curve). Large b means the same trade barely moves price (flat curve). Maximum market maker loss also scales with b (for n outcomes, worst-case on the order of b · ln(n) in classic LMSR).
Operators pick b balancing trader experience (low slippage), capital at risk (subsidy cap), and fee income to recoup subsidy. b decides how many dollars of flow must arrive to change the public probability—so it controls how noisy the headline percent is.
Slippage intuition
Same YES buy, different b: with b = 10, a $500 buy might jump several percentage points and a $5,000 buy can reprice ten percent or more—manipulation cheap. With b = 100, moves are moderate for small accounts. With b = 1000, even five-figure flow may barely nudge the curve. Slippage is the price of instant liquidity from a finite pool, not a bug.
Worked example: three trades, same b
Market starts 50% / 50%, binary LMSR, b = 100. A $200 YES buy might lift implied probability toward 51%—noise-sized. An $800 buy might reach 54%—still reasonable for retail. A $3,000 buy might print 61%—you moved the market. Journalists citing “68%” after one whale clip are often citing b, not consensus.
If b = 25 instead, that $3,000 clip might print 68%+ from the same information—same flow, wildly different public odds.
Liquidity is not infinite
Large b feels deep, but someone prefunded that depth. Capital efficiency asks how many dollars of subsidy per dollar of tradable size. Retail sees tight one-percent moves; treasury sees locked collateral. Operators and LPs bet fees plus volume repay a high b; regulated venues bet professional makers supply depth without locking entire notional in one pool.
Tuning mistakes
b too low for a hype event whips headline odds on memes, makes manipulation cheap, and hurts LPs. b too high for long-tail ties capital on contracts with $200 volume—bad ROI. b static while volatility rises leaves LPs exposed when book makers would widen spread—arb bots may drain the pool while percent lags the regulated touch.
Relating b to order-book spread
Rough analogy (not exact): small b ≈ wide spread plus thin depth; large b ≈ tight spread plus deep book. Difference: AMM slippage is deterministic from formula; CLOB spread is strategic from makers who pull quotes in crises.
| Feature | CLOB touch | LMSR with b |
|---|---|---|
| Quote after bad news | Makers widen spread | Curve moves; slippage rises |
| Who adjusts | Human or algo discretion | Math unless hybrid |
| Transparency | Visible ladder | Disclosed b + TVL + simulator |
Estimating trade impact (practice)
Before a $2,000 YES buy: note current implied p and b if disclosed; use the platform simulator; if price moves more than two to three percent for your size, treat edge as slippage-adjusted; compare to book VWAP on the same event elsewhere.
Example: book YES ask 58¢ with $4k stacked to 60¢—your $2k VWAP might be 59¢. Pool mid 56% but simulator 59.2% effective on $2k with b = 120. The AMM looked cheaper at mid but more expensive all-in—arb may close the gap, but takers cannot assume it.
Election-week dynamics
During high-volatility windows, a fixed b cannot widen spread the way human makers do on a book. LPs may withdraw, leaving a large b pool with fewer counterparties—percent moves faster than usual even though the parameter did not change. Hybrids exist partly to restore discretionary spread widening at the touch while keeping the pool as backstop.
Platform disclosure
Good venues publish b or equivalent liquidity metric, pool TVL, and historical volume. Bad venues show percent only—treat as marketing. Regulated CLOBs publish order book depth instead of b because they are not LMSR-first. When comparing venues, translate book depth at touch into b + TVL + simulator.
Hybrids
Hybrids may use b for baseline and CLOB for overlay—effective depth is pool plus book, not b alone. A large b with active makers means the curve is insurance; your short-term price may still be the book.
Tie-in to informative prices and manipulation
Informative prices need real aggregation. Low b markets can show dramatic percent moves with little informational content—violating the spirit of crowd prices even when math is correct. Manipulation cost scales with cost to move displayed odds: small b lowers that cost on pools; on books, manipulators must lift visible size at multiple levels. A sudden twelve-point jump on a crypto prop may be information—or b too small.
Operators launching niche props might start conservative and raise b only if volume justifies lockup; election marquees often prefer hybrid or CLOB rather than b alone.
Dynamic b (when venues offer it)
Some protocols adjust b or equivalent depth parameters from volume or volatility. Traders should read release notes: a market that felt “deep” last month may be shallow today because parameters changed. Static thinking about liquidity causes surprise slippage after governance votes.
Comparing b across venues
There is no universal mapping from b = 100 on platform A to b = 100 on platform B if fee schedules and outcome token mechanics differ. Compare simulated impact for your notional instead of raw b when choosing where to trade.
Worked sensitivity narrative
Hold event belief fixed at 60% true chance. With b = 40, a $1,000 YES buy might display 64% afterward—you moved the public number four points. With b = 400, the same buy might show 60.5%. Neither move is “information” unless the flow was informed; both are mechanical. Your job is to separate signal trades from impact trades when reading headlines after large clips.
Operator recap
Think of b as a volume knob on how loud the market shouts per dollar. Turn it up for marquees only when fee math covers the subsidy; turn it down for experiments; never treat b alone as marketing depth without TVL and volume beside it.
If disclosed b is missing, infer sensitivity from a $500 and $2,000 simulator probe—the curve tells you more than the tile.
Summary sentence
b is the platform’s answer to “how many dollars must trade before the public percent moves”—know that answer before you trust the percent.
Key ideas
b sets slippage sensitivity and caps subsidized loss. Higher b needs more locked capital for marginal improvement. Stress-test your size against the curve, not headline mid. Pair b with TVL and volume—never read probability in isolation.
Next: AMM vs order book—decision framework.