Predyx Market Creation Guide (Plain English)

TL;DR

• Predyx uses LMSR (a type of automated market maker) to set prices and pay out.
• You, as the creator, choose a liquidity setting (called b) and a fee. These two choices control both how stable your prices are and your risk/reward.
• Your maximum possible loss as the market’s liquidity provider is called WCL (Worst-Case Loss). In Predyx we treat the initialCost you put in as covering this WCL.
• WCL grows with b and with the number of outcomes.
• Fees earned from trading can offset or fully cover your WCL. Break-even volume is roughly WCL ÷ fee_rate.
• Today you can create: Binary (YES/NO) and Range (Multi-option) markets. More types are coming (see roadmap at the end).

1) What you can create today

• Binary (YES/NO): Two outcomes. Simple and popular. Examples: “Team A wins?” “Will BTC close above 120k on Oct 31?”
• Range (Multi-option): Several discrete options that cover a scale or set of ranges. Example: “BTC closing price on Friday: 100–105k, 105–110k, 110–115k, …”

On the roadmap (priority order): Set-markets → Scalar markets (numeric & date) → Auto-resolving sports markets with live scores.

2) LMSR in one minute

• When people buy shares in an outcome, its price goes up a bit.
• Prices always sum to 100% across all outcomes.
• Your liquidity parameter b controls how much prices move when people trade. Bigger b = steadier prices, but also higher risk cap (WCL) for you, the market creator.

Worst-Case Loss (WCL) = b × ln(number_of_outcomes)

3) Key terms

• initialCost: The sats you commit when creating the market. On Predyx, this funds the market maker. Treat it as your risk budget.
• WCL (Worst-Case Loss): The most you can lose from the market maker, even in the most extreme outcome.
• b (liquidity): Your “stability” knob. Higher b → smoother prices → higher WCL.
• Fee (market fee): % charged on trades. This is your revenue stream to offset WCL.
• Outcomes: The available choices (2 for binary; many for range).

On Predyx: initialCost is there to cover WCL. If trading volume is low and fee income is small, you may not earn back your initialCost.

4) WCL, number of outcomes, and initial odds

WCL = b × ln(1 / p_min)
where p_min is the smallest initial probability among all outcomes.

If all outcomes are equal: p_min = 1/n ⇒ WCL = b × ln(n)

• n = 2 → ln(2) ≈ 0.6931
• n = 3 → ln(3) ≈ 1.0986
• n = 4 → ln(4) ≈ 1.3863
• n = 5 → ln(5) ≈ 1.6094
• n = 10 → ln(10) ≈ 2.3026

• Adding more outcomes increases WCL (even odds).
• Skewed initial odds also affect WCL—what matters is the smallest starting probability. A very tiny p_min (e.g., a long-shot bucket) can dominate WCL.

Call-out on “favorites”: Setting a heavy favorite increases WCL (because p_min gets smaller) and usually requires more initialCost to tilt prices. But if the favorite actually wins, your realized loss often ends up lower, and fees can push you net-positive. Plan using WCL; hope for better.

5) Fees offset WCL (break-even math)

• Fee revenue ≈ fee_rate × total_traded_volume.
• If you ignore any leftover inventory value at the end (safe, conservative assumption), then a simple break-even is:

Break-even volume ≈ WCL ÷ fee_rate

Example A — Binary (YES/NO)

• Choose b = 100,000 sats
• WCL ≈ 0.6931 × 100,000 = 69,310 sats (rounding okay)
• Fee = 2% (0.02)
• Break-even volume ≈ 69,310 ÷ 0.02 = 3,465,500 sats

Example B — Range (5 outcomes)

• Choose b = 100,000 sats, n = 5
• WCL ≈ 1.6094 × 100,000 = 160,940 sats
• Fee = 3% (0.03)
• Break-even volume ≈ 160,940 ÷ 0.03 ≈ 5,364,700 sats

Note: In practice, you may have positive or negative leftover inventory when the market resolves. That can shift the actual result. The break-even formula above is a simple, conservative planning tool.

6) Picking your b (liquidity) and fee: a quick playbook

• New or niche topic? Lower b (e.g., 50k–100k sats) so prices can move and attract interest without over-committing WCL.
• Hot/evergreen topic? Higher b (e.g., 150k–300k sats) for steadier prices when you expect lots of trading.
• Range markets: Remember b impacts every outcome; since WCL scales with ln(n), don’t over-inflate b when you have many options.

• Crowd-pleasers with volume: 1%–2% can be enough; volume covers WCL.
• Specialty/low-volume: 2%–4% to ensure fees meaningfully offset WCL.
• Promos: You can go low at launch (even ~0.5%–1%) to spark activity—but your break-even volume jumps. Budget for that.

Sanity check: Estimate your WCL first, then confirm you’re comfortable with the break-even volume = WCL ÷ fee_rate.

7) Starting prices & initial odds (super important)

• Uniform start (e.g., 50/50 for YES/NO; equal ranges) keeps WCL at b × ln(n).
• Skewed start (heavy favorite): p_min gets smaller → WCL increases. You’ll also spend more initialCost to push prices away from even.
• However, if your prior is accurate (the favorite actually wins), your realized loss often ends up lower than worst-case, and fees may put you net-positive. Worst-case planning is still essential.

1. Collect lines from multiple reputable sportsbooks and other prediction markets for the same event.
2. Convert to implied probabilities (remove the book’s margin if possible).
3. Average or use the median of those probabilities as your starting odds.
4. Avoid extreme skews unless you expect strong volume/fees—remember p_min drives WCL.

• Uniform 50/50: p_min = 0.50 → WCL ≈ 0.6931 b = 69,310 sats
• 70/30 favorite: p_min = 0.30 → WCL ≈ ln(1/0.30) × b ≈ 1.2040 b = 120,400 sats
• 90/10 favorite: p_min = 0.10 → WCL ≈ ln(10) × b ≈ 2.3026 b = 230,260 sats

• 50/50 → ≈ 3.47M sats
• 70/30 → ≈ 6.02M sats
• 90/10 → ≈ 11.51M sats

8) Risk: when you might lose initialCost/WCL

• If the market doesn’t attract enough trading volume, your fees may not cover WCL.
• If traders push strongly to the winning outcome, your realized loss can approach WCL (that’s why we plan around it).
• If your fee is set too low for the topic’s actual volume, you may finish negative on the market maker.

• Pick a reasonable b (don’t over-commit).
• Use clear, popular topics and share your market to drive volume.
• Set fees in line with expected interest.
• Keep outcome count focused (especially for range markets).

9) Resolution criteria: keep it crystal-clear

Resolve to YES if: [exact event/condition]. Otherwise NO.
Deadline: [date & time, with timezone].
Primary source: [official site or data feed].
If the event is canceled/unclear: [state fallback: N/A / Other / No].
Allow credible consensus if primary source is down.

• Price markets: “Resolves by the [exchange/symbol] official close on [YYYY-MM-DD]. Source: [link/name].”
• Sports: “Resolves to the official result posted by [league/site]. If postponed beyond [date], resolves to N/A.”
• Politics: “Resolves based on official certification (e.g., Congress / election commission).”

10) Practical guardrails & common pitfalls

• Too-large b on a thin market → big WCL, not enough fees to cover it.
• Too many outcomes in a range → high WCL, thin activity per option.
• Ambiguous criteria → disputes and unhappy traders.
• Fees too low for niche topics → slow fee recovery.
• Unclear closing time → last-minute confusion. Always set close and resolution times explicitly.

11) Worked mini-scenarios (in sats)

Scenario 1: Binary, medium interest

• b = 120,000 → WCL ≈ 0.6931 × 120,000 = 83,172
• Fee = 2% → Break-even volume ≈ 83,172 ÷ 0.02 = 4,158,600
• Expect 6–8M sats of volume? Fine. If it actually does 6M sats, fee ≈ 120,000, which should more than cover WCL.

Scenario 2: Range (4 options), moderate interest

• b = 90,000, n = 4 → WCL ≈ 1.3863 × 90,000 = 124,767
• Fee = 3% → Break-even ≈ 124,767 ÷ 0.03 = 4,158,900
• If you only expect ~2M sats of volume, either raise fee (e.g., 4%) or reduce b (e.g., 60k) or cut outcomes to 3.

12) Roadmap (what’s coming next)

• Set-markets (choose many items that may all be true)
• Scalar markets (numeric results and date/time results with auto-grading)
• Auto-resolving sports with live scores

13) Quick glossary

• Share: A claim on payout if that outcome happens.
• Price: The current cost to buy one share of an outcome.
• b (liquidity): Controls how sensitive prices are to trades (and the WCL size).
• WCL: Your maximum possible loss as the market’s liquidity provider. WCL = b × ln(n) for even starts; WCL = b × ln(1/p_min) for skewed starts.
• initialCost: The sats you commit upfront to fund the LMSR; practically, treat it as your WCL budget.
• Fee: % charged on trades, your primary way to earn back WCL.

14) Creator checklist (copy/paste)

• Pick market type (Binary / Range)
• Write a clear, unambiguous resolution criteria (include primary source + timezone)
• Confirm the market end/close date & time are accurate (timezone set correctly)
• Choose a short, scannable title
• Add an attractive 1:1 thumbnail (square) that matches the market
• Set initial odds using reputable sportsbooks & other prediction markets (de-vig when possible). Avoid extreme skews; remember WCL = b × ln(1/p_min).
• Choose number of outcomes (keep it tight)
• Select b and compute WCL (uniform: b × ln(n); non-uniform: b × ln(1/p_min))
• Choose fee and compute break-even volume = WCL ÷ fee_rate
• Sanity-check: Are you comfortable with that volume goal?
• Set close time and resolution time explicitly (if not already done)
• Publish, share the link, and monitor activity

15) Final tips

• Start with binary markets to learn the feel of b, fees, and volume.
• Prefer fewer, clearer outcomes in range markets.
• If in doubt, lower b and add volume before increasing it later.
• Use attractive thumbnails and short titles to boost engagement.
• Keep experimenting—LMSR is forgiving when you plan around WCL and fees.

16) Vig remover (de-juicing) cheat sheet

Step-by-step (works for 2-way and 3-way markets)

1. Convert odds → implied probabilities
   • American +A → p = 100 / (A + 100)
   • American −A → p = |A| / (|A| + 100)
   • Decimal D → p = 1 / D
2. Add them up: S = Σ p_i. With vig, S > 1.
3. Remove the vig (normalize): p_i^fair = p_i / S. Now the fair probs sum to 1.
4. (Optional) Convert back to odds
   • Probability p < 0.5 → American +100 * (1/p − 1)
   • Probability p ≥ 0.5 → American −100 * p / (1 − p)
   • Probability p → Decimal 1 / p

Tip: Pull prices from multiple sportsbooks & prediction markets, average the implied probabilities, then de-juice once.

Example A — Two-way moneyline

• Given: Team A −170, Team B +150
• Implied (raw):
  p_A = 170 / (170 + 100) = 0.6296, p_B = 100 / (150 + 100) = 0.4000
  Sum S = 1.0296
• Fair:
  p_A^fair = 0.6296 / 1.0296 = 0.6115 → American ≈ −157
  p_B^fair = 0.4000 / 1.0296 = 0.3885 → **American ≈ +157`

Example B — Three-way (soccer)

• Given: Home +120, Draw +230, Away +220
• Implied (raw):
  Home = 100/220 = 0.4545, Draw = 100/330 = 0.3030, Away = 100/320 = 0.3125
  Sum S = 1.0701
• Fair:
  Home 0.4248 (≈ +135), Draw 0.2832 (≈ +253), Away 0.2920 (≈ +242)

Use these fair probabilities as your initial odds. Remember: a very small p_min (tiny long-shot bucket) will increase WCL via WCL = b × ln(1/p_min).

17) Quick probability ↔ odds converter

• Prob → Decimal: D = 1 / p
• Prob → American:
  • if p < 0.5: +100 * (1/p − 1)
  • if p ≥ 0.5: −100 * p / (1 − p)
• American → Prob:
  • if +A: 100 / (A + 100)
  • if −A: |A| / (|A| + 100)

Workflow: Pull several books/markets → convert to probs → average → de-juice → use as initial odds in Predyx. Keep an eye on p_min to manage WCL and your initialCost.