AI Edge: The Numbers Say Charles Oliveira Offers Hidden Value

This is a hand-weighted logistic model (not a full ML model) so you can see exactly how each factor changes the prediction.

Features used (Oliveira minus Gamrot):

• Strikes landed per minute (SLpM)

• Significant strikes absorbed per minute (SApM)

• Takedown rate (avg per 15 minutes)

• Takedown defense (percentage)

• Submission attempts per 15 minutes

• Home advantage (Oliveira is in Brazil)

  
  

I normalized as simple differences (Olive minus Gamrot) and used reasonable domain weights based on how important each skill usually is in this matchup (substrate: grappling-heavy match where submissions and takedown control matter most).

Weights used (you can tweak these):

• SLpM: 0.30

• SApM: -0.15

• Takedown avg: -0.20

• Takedown defense: 0.02 (per percentage point)

• Submission avg: 0.40

• Home: 0.15

Using the stats above, the model computed a score and converted it to probability with a logistic (sigmoid) function.

Contributions to Oliveira’s score (positive helps Oliveira, negative hurts):

• SLpM (Oliveira > Gamrot): +0.015

• SApM (Oliveira absorbs more): −0.025

• Takedown avg (Olive fewer takedowns): +0.620  (note: negative sign in formula turns a higher Gamrot takedown rate into a penalty for Oliveira; numeric sign here reflects how I structured the diffs/weights)

• Takedown defense (Oliveira much lower than Gamrot): −0.680

• Submission avg (huge Oliveira edge): +0.996

• Home advantage (Brazil): +0.150

Total model score = ≈ 1.076 -> convert with sigmoid -> Oliveira win prob = 0.746 (74.6%)

I ran a 20,000-trial Monte Carlo using the heuristic model we discussed, adding reasonable fight-to-fight randomness to each fighter stat. Here are the results and what they mean.

Monte Carlo summary (20,000 trials)

• Mean model probability (Oliveira): 0.7159 (71.6%)

• Median model probability: 0.7303 (73.0%)

• Simulated Oliveira win % (Bernoulli sampling): 71.37%

• 10th–90th percentile of model p: 0.5634 – 0.8474

• Per-trial probability SD: 0.1107

A histogram of the per-trial model probabilities was produced (distribution peaked around ~0.70–0.80; notable tail down into the 0.4–0.6 range when sampled stats swing against Oliveira).

Quick interpretation

• The Monte Carlo confirms a clear lean toward Charles Oliveira in this heuristic model — roughly ~71–72% chance to win.

• The 10th–90th percentile band (≈56% to 85%) shows uncertainty: if you believe Gamrot’s wrestling/TD control will dominate on fight day, that scenario is well within the model’s plausible outcomes.

• The simulated win percentage (71.37%) is the fraction of simulated fights where, after sampling variability, Oliveira “won” when we used the model’s per-trial probability to decide the outcome.

Why this differs slightly from the single-point heuristic number you saw earlier (~74.6%)

• The earlier value was a single deterministic computation using mean stats. The Monte Carlo adds realistic variability (random sampling around each stat), which pulls the average model probability down a bit — because when Gamrot’s takedown numbers sample high and Oliveira’s submission attempts sample low, the matchup tightens. That variability is exactly why a Monte Carlo is useful.

I converted the Monte-Carlo output into no-juice (fair) betting odds in three common formats. I did the conversions for both the model mean probability (from the per-trial model p) and the simulated win % (the fraction of simulated fights that resulted in an Oliveira win). Both are close — I’ll show them both so you can pick which you prefer to use.

Input probabilities

• Mean model probability (average per-trial p): 0.7159 (71.59%)

• Simulated win % (Bernoulli sampling): 0.7137 (71.37%) Conversions below assume no bookmaker

  margin (fair odds = 1 / probability).

• 
  

Fair odds (from 0.7159)

• Decimal (European): 1.397 (≈1.39684…)

  • Interpretation: a $1 bet returns $1.3968 total (profit $0.3968) if Oliveira wins.

• Fractional (British): ~0.397/1 → roughly 2/5 in common betting shorthand.

  • Interpretation: risk 5 to win 2 (favorites are often shown as fractions <1).

• American (moneyline): −252

  • Interpretation: you must bet $252 to win $100 (Oliveira is the favorite).

    
    

Fair odds (from 0.7137)

• Decimal: 1.401 (≈1.40115)

• Fractional: ~0.401/1 → roughly 2/5 (very similar)

• American: −249

Quick math (assuming ML −102 is for Oliveira)

• Market implied probability for −102 = 102 / (102 + 100) = 0.50495 → 50.5%.

• Model probability (mean) = 0.7159 (71.59%).

• Model − Market gap = 0.7159 − 0.50495 = 0.21095 → ≈21.1 percentage points of edge for Oliveira.

  
  

Fair EV at −102

• Decimal odds for −102 = 1 + 100/102 ≈ 1.98039.

• Expected return per $1 using model p = p decimal − 1 = 0.7159 1.98039 − 1 ≈ +$0.4178 → +41.8% ROI.

• So a $100 bet has an expected profit ≈ $41.78 under the model.

  
  

  

  
  

  I ran a sensitivity sweep varying the takedown weight (TD_avg) from −0.40 to −0.05 and the submission weight (SUB_avg) from 0.00 to 0.80, using Monte Carlo trials at each grid point (3,000 trials per point). That produced a grid of estimated Oliveira mean win probabilities across those weight combinations and a heatmap (shown).

  
  

  Here are the key takeaways and how to read the results.

  
  

  Quick summary (what the sweep shows)

  • Across the entire grid, Oliveira’s mean win probability ranges roughly from ~0.45 (worst-case-ish grid corners) up to ~0.90 (when submissions are weighted very high and takedowns are weighted lightly).

  • With weights near our original model (SUB ≈ 0.40, TD ≈ −0.20) the mean Oliveira probability sits solidly in the ~0.70–0.76 range (consistent with earlier Monte Carlo).

  • Even in fairly pessimistic scenarios (where takedowns matter a lot: TD weight ≈ −0.40, and submissions matter little: SUB weight ≈ 0.0–0.2), Oliveira still frequently stays above the market-implied 50.5% — though there are grid points where the model drops to around market or below. The exact count of grid points ≤ market was printed in the output (you can inspect the table).

  • Best-case for Oliveira (SUB high, TD low): probabilities ~0.85–0.90.

  • Worst-case for Oliveira (SUB low, TD very negative): probabilities can fall into the mid-40s–50s — where the market price would be fair or even favorable to Gamrot.

The red contour line marks where the model’s implied win probability equals the market-implied 50.5% for Oliveira (−102).

• The region above/right of the line (greener zone) → model says Oliveira is undervalued (positive expected value).

• The region below/left of the line (darker zone) → Gamrot has the edge under those extreme assumptions (high takedown weight, low submission weight).

In plain English: unless you believe takedowns dominate and submissions barely matter, the model favors Oliveira at −102.

Use our link below to get up to $300 in free bets INSTANTLY by placing a $5 bet in MO!