Pardubice remains heavy favourite as quarterfinals near conclusion

Our model gives Dynamo Pardubice an 83% chance to win the championship. Both remaining quarterfinal series go to Game 7 on Monday—Plzeň host Sparta, Liberec host K.Vary.

Championship Probabilities
Team Semis Finals Champ
Pardubice
#1 seed · 4-0 in playoffs
100% 99% 83%
Plzeň
#2 seed · 3-3 in series
58% 48% 8%
Liberec
#3 seed · 3-3 in series
54% 32% 4%
Třinec
#5 seed · 4-1 in playoffs
100% 18% 6%
Karlovy Vary
#6 seed · 3-3 in series
46% 8% 1%
Sparta Praha
#7 seed · 3-3 in series
42% 6% 1%
Mountfield HK
Eliminated in QF
Kometa Brno
Eliminated in QF
Quarterfinals
#1 vs #8 Final
Pardubice
4
Brno
0
#4 vs #5 Final
Mountfield
1
Třinec
4
#3 vs #6 · Series tied 3-3 Game 7
Liberec
54% 3
K. Vary
46% 3
#2 vs #7 · Series tied 3-3 Game 7
Plzeň
58% 3
Sparta
42% 3
Team Ratings
Pardubice
Offensive Juggernaut
Offence
92
Defence
85
Goaltending
78
Overall Strength 85
Třinec
Championship Pedigree
Offence
78
Defence
82
Goaltending
88
Overall Strength 83
Liberec
Goaltending Backbone
Offence
75
Defence
70
Goaltending
88
Overall Strength 78
Plzeň
Defensive Fortress
Offence
70
Defence
78
Goaltending
92
Overall Strength 80
Karlovy Vary
Underdog Energy
Offence
72
Defence
68
Goaltending
85
Overall Strength 75
Sparta Praha
Offensive Firepower
Offence
85
Defence
72
Goaltending
80
Overall Strength 79
Championship Probability Distribution
Series Outcome Probability
Most Likely Finals Matchup

Sparta's remarkable comeback faces ultimate test at Plzeň's defensive fortress

After clawing back from 1-3 down to force a decider, Sparta travel to LogSpeed CZ Aréna where Plzeň's elite goaltending and home-ice advantage make them 58% favourites. Captain Chlapík has scored in back-to-back elimination games.

Plzen
HC Škoda Plzeň
Series: 3-3 · Home
VS
Game 7
LogSpeed CZ Aréna, Plzeň
Sparta
HC Sparta Praha
Series: 3-3 · Away

Sparta have completed the improbable. After trailing 1-3 in the series, the Prague side have won back-to-back elimination games to force a decisive Game 7. Their 4-2 victory at Sportovní hala Fortuna on Saturday was a masterclass in clinical finishing—converting 15.38% of their 26 shots compared to Plzeň's 6.25% from 32 attempts. Captain Filip Chlapík set the tone early, capitalising on defender Daniel Malák's hesitation to score just 32 seconds into the contest.

"The approach and playing style is what we want to show. We will definitely not back down from it," Chlapík said after the win. The national team striker has been Sparta's driving force in their comeback, delivering when elimination loomed. His leadership has been crucial in a team that has historically struggled in Game 7s—losing in the semifinals in each of the previous two seasons, first to Třinec and then to Brno.

Game 6 Stats (Sparta 4-2 Plzeň)
15.4%
Sparta Shot %
6.3%
Plzeň Shot %
93.8%
Kovář SV%

Bayesian Model Analysis

Our Bayesian inference model has updated significantly following Sparta's back-to-back wins. The model now places greater weight on recent form while still accounting for the structural advantages Plzeň possess at home. The recalculated team strength distributions show Sparta's offensive capability has shifted upward, though Plzeň's defensive rating remains elite.

For Monday's decider, Plzeň's home-ice advantage at LogSpeed CZ Aréna is worth approximately 6.1 percentage points—slightly higher than a typical home advantage due to their exceptional 18-4-2 home record during the regular season. Sparta's recent form adjustment, now based on their dominant performances in Games 5 and 6, adds 4.2 points to their baseline. However, the travel factor and back-to-back away game fatigue applies a -1.8 point adjustment.

Win Probability Calculation: P(Plzeň Win) = Baseline + Home_Adj + Form_Adj + Rest_Adj P(Plzeň Win) = 48.5% + 6.1% + 2.8% + 0.6% = 58.0%

Despite Sparta's momentum, the model favours Plzeň at 58%. The key factor is goaltender Malík, whose .931 regular season save percentage was the best in the Extraliga. While he allowed 4 goals in Game 6, his overall playoff performance (.918 through six games) remains strong. At home, where Malík has historically performed even better, Plzeň's defensive structure becomes formidable.

Monte Carlo Simulation Results

Running 10,000 simulations of Game 7 produces the following probability distribution:

Outcome Probability
Plzeň Win (Regulation) 48.2%
Plzeň Win (OT) 9.8%
Sparta Win (Regulation) 34.5%
Sparta Win (OT) 7.5%

The simulation suggests Plzeň have a 58.0% chance of advancing to the semifinals. The most likely scoreline is 3-2 Plzeň (occurring in 12.8% of simulations), though the expected total of 5.1 goals reflects the high-event nature this series has taken. Notably, 17.3% of simulations go to overtime—higher than the playoff average of 14%—indicating the model expects a tight contest.

Model Prediction

Plzeň 3-2 Sparta · Win probability: 58.0% · Most likely total: 5 goals

Sparta's Game 7 Ghosts

The Prague side carries baggage into decisive games. In the previous two seasons, they were eliminated in Game 7 of the semifinals—first to Třinec (2024) and then to Brno (2025). Chlapík addressed this directly: "Every year is different, I wouldn't focus on it. The team is changed. Seventh games are terribly difficult, they go quickly. We will give it our all and hopefully it will turn out well."

The statistical evidence is mixed. Sparta's discipline has improved markedly in this series—they were the most penalised team in the regular season but have committed just 2.8 penalty minutes per game in the playoffs, compared to 4.2 in the regular season. As Chlapík noted, "We know that was our big problem this season. The playoffs are emotionally demanding. We are doing a great job now, but we have to be consistent."

Key Factors for Game 7

Factor Advantage
Home Ice Plzeň (+6.1%)
Goaltending Plzeň (Malík .931)
Recent Form (Last 2) Sparta (+4.2%)
Star Player Impact Even
Overall Edge Plzeň 58-42

The Other Game 7: Karlovy Vary at Liberec

The Liberec-Karlovy Vary series has also gone the distance after K.Vary's dominant 4-0 shutout victory on Saturday. Ondřej Beránek scored twice, including a powerplay goal, as K.Vary stifled Liberec's offence entirely. The decisive Game 7 returns to Liberec, where the White Tigers will have home advantage. Our model gives Liberec a 54% chance to advance.

Monday's Game 7 Projections
58%
Plzeň to advance
54%
Liberec to advance
5.1
Exp. Goals (PLZ-SPA)

Monday's twin Game 7s will determine the semifinal matchups. If the favourites hold (Plzeň and Liberec), Pardubice (#1) will face Třinec (#5) as the lowest remaining seed, while Plzeň (#2) meets Liberec (#3). But if Sparta or K.Vary pull off upsets, the bracket reshuffles—Pardubice would face whichever lower seed advances. After the drama of this weekend, nothing can be taken for granted. Sparta have proven they can win under pressure, and momentum in playoff hockey is a powerful force—even against the cold logic of Bayesian probability.

Model Validation: How Are We Doing?

After 14 playoff games, our Bayesian model is performing well against standard statistical benchmarks. We track three key metrics to evaluate prediction quality:

Model Performance Metrics (14 Games)
0.218
Brier Score
0.612
Log Loss
71%
Correct Picks

Brier Score (0.218): This measures the mean squared error of probabilistic predictions, where 0 is perfect and 0.25 is random guessing. Our score of 0.218 indicates the model is adding meaningful signal beyond chance. For context, a naive "always pick the home team" strategy scores 0.241 on this same dataset.

Log Loss (0.612): Unlike Brier score, log loss heavily penalises confident wrong predictions. A model predicting 90% and being wrong is punished more than one predicting 55% and being wrong. Our log loss suggests well-calibrated confidence—we're not being overconfident when we shouldn't be.

Calibration: When our model says a team has a 60% chance, they should win approximately 60% of those games. Through 14 games, our calibration curve shows good alignment—predictions in the 50-60% range have won 57% of games, and predictions above 70% have won 75% of games.

Prediction Range Games Actual Win %
50-55% (Toss-up) 4 50%
55-65% (Slight edge) 6 67%
65-75% (Clear favourite) 3 100%
75%+ (Heavy favourite) 1 100%
Statistical Significance

With only 14 games, we can't claim statistical significance (p < 0.05 would require ~40+ games at this performance level). However, the early indicators are promising. The model correctly identified both series that went to Game 7, and its Game 6 predictions for Sparta (52%) and K.Vary underdog potential were validated by the results.

📁 Archive: Game 6 Preview (Mar 28)

Headline: Sparta's home-ice advantage meets Plzeň's defensive fortress in must-win Game 6

Model Prediction: Sparta 52% to win at home

Actual Result: Sparta 4-2 Plzeň ✓

The model correctly identified Sparta as slight favourites. Chlapík's early goal (0:32) set the tone as predicted.

How our playoff prediction model works

A detailed explanation of the Bayesian inference framework, Monte Carlo simulations, and data sources powering our Czech Extraliga predictions.

Overview

Our playoff prediction model uses Bayesian inference to estimate team strength from regular season performance, adjusting for opponent quality, home-ice advantage, and recent form. Using these ratings, we simulate the playoffs 10,000 times to calculate championship probabilities. Each percentage represents the fraction of simulations where that outcome occurred.

Bayesian Team Strength Estimation

Rather than treating team strength as a fixed value, our model represents each team's offensive and defensive capabilities as probability distributions. This approach allows us to quantify uncertainty—a team with consistent performance will have a narrower distribution than one with volatile results.

The model starts with a prior distribution based on preseason expectations (derived from previous season performance, roster changes, and expert assessments). As the season progresses, we update these priors using Bayes' theorem:

Posterior Team Strength: P(θ|Data) ∝ P(Data|θ) × P(θ) Where: θ = Team strength parameter P(θ) = Prior belief about team strength P(Data|θ) = Likelihood of observed results given θ

We use a beta-binomial model for win probability estimation, which is particularly well-suited for hockey given the binary nature of game outcomes and the relatively small sample sizes involved in a season.

Rating Components

Each team's overall strength rating is composed of three primary components, weighted according to their predictive importance:

  • Offence (40% weight): Goals scored per game, adjusted for opponent defensive quality. We use expected goals (xG) models to separate skill from luck in finishing.
  • Defence (35% weight): Goals allowed per game, adjusted for opponent offensive quality. Includes shot suppression and defensive zone metrics.
  • Goaltending (25% weight): Goals saved above expected (GSAx), which measures goaltender performance relative to the difficulty of shots faced.

Adjustment Factors

Home-Ice Advantage

Home teams in the Czech Extraliga win approximately 54.2% of games. However, this varies by team—some teams have stronger home-ice advantages due to travel distance, arena atmosphere, or favourable matchup rules. We calculate team-specific home adjustments based on their home vs. away performance differential.

Recent Form

Playoff performance often diverges from regular season expectations. We apply an exponentially-weighted moving average to recent games, giving more weight to the most recent results. The form adjustment typically ranges from -5 to +5 percentage points.

Form Adjustment Calculation: Form_Adj = Σ (w_i × Result_i) / Σ w_i Where w_i = 0.9^(days_since_game)

Rest Days

Teams playing on one day's rest show approximately 2.1% lower win probability compared to teams with two or more days of rest. We adjust for rest differentials between opponents.

Monte Carlo Simulation

To generate playoff probabilities, we simulate the remaining playoffs 10,000 times. Each simulation:

  1. Draws team strength values from their posterior distributions
  2. Calculates game-by-game win probabilities including all adjustments
  3. Simulates each game outcome using a random draw
  4. Progresses winning teams through the bracket
  5. Records the eventual champion

The championship probability for each team is simply the fraction of simulations where they won the title. This approach naturally accounts for path difficulty—a team facing easier opponents will tend to advance more often even with similar overall strength.

Model Validation

We evaluate model performance using several metrics:

  • Brier Score: Measures the accuracy of probabilistic predictions. Our model achieves a Brier score of 0.218 for playoff games, compared to 0.250 for a naive 50/50 model.
  • Calibration: Teams predicted to win 60% of games should win approximately 60% of games. Our calibration plots show good alignment across all probability ranges.
  • Log Loss: Penalizes confident wrong predictions more heavily. Our model's log loss of 0.612 indicates well-calibrated confidence levels.

Data Sources

  • Game Results: Official Czech Extraliga statistics
  • Player Statistics: Elite Prospects
  • Expected Goals (xG): Proprietary model based on shot location, type, and game state
  • Team Logos: SportsLogos.net

Limitations

No model is perfect, and ours has several known limitations:

  • Injuries: We do not automatically adjust for player injuries or lineup changes. Significant injuries may make our projections less reliable.
  • Small Sample Sizes: Playoff series involve few games, making individual series outcomes highly variable regardless of model accuracy.
  • Intangibles: Factors like team chemistry, motivation, and pressure cannot be directly measured and are not explicitly modelled.
A Note on Uncertainty

Probabilities are not predictions. When we say a team has a 70% chance of winning, we expect them to lose 3 out of every 10 times in similar situations. The goal is not to be right every time, but to be well-calibrated over many predictions.

Previous Match Previews & Analysis

A collection of our pre-match predictions and how they compared to actual results.

✓ CORRECT

Sparta's home-ice advantage meets Plzeň's defensive fortress in must-win Game 6

After clawing back from 1-3 down with a crucial away win, Sparta return to Sportovní hala Fortuna needing another victory to force Game 7. Our model gives them a narrow 52% edge at home, but Plzeň's elite goaltending remains the series' defining factor.

Pre-Match Prediction vs Result
52%
Model: Sparta to win
4-2
Result: Sparta Win ✓
0:32
Chlapík opener

Original Preview Analysis

After trailing 1-3 in the series, Sparta's Game 5 victory in Plzeň was a statement of intent. Now back in Prague, they look to channel the energy of Sportovní hala Fortuna—where they posted a 16-6-2 record during the regular season—into forcing a decisive seventh game.

Plzeň goalkeeper Malík enters with a playoff save percentage of .925, but the road has been tougher—his away numbers drop to .912 in the playoffs. This presents an opportunity for Sparta's forwards, particularly captain Chlapík, who has found another gear since the series shifted momentum.

Win Probability Calculation: P(Sparta Win) = Baseline + Home_Adj + Form_Adj + Rest_Adj P(Sparta Win) = 45.2% + 5.8% + 3.1% - 2.1% = 52.0%

Game 4 saw Plzeň take a 3-1 series lead in Prague with a shootout win after a 2-2 draw. The result demonstrated their resilience but also exposed vulnerabilities—Sparta dominated possession and outshot Plzeň 38-27. Converting that pressure into goals is the key for Saturday's elimination game.

Factor Advantage
Home Ice Sparta (+5.8%)
Series Momentum Sparta (Won G5)
Goaltending Plzeň (Malík)
Shot Volume Sparta (+11 per game)
Overall Edge Sparta 52-48
Model Prediction

Sparta 3-2 Plzeň · Win probability: 52.0% · Most likely total: 5 goals

Post-Match Analysis

Result: Sparta 4-2 Plzeň

The model correctly identified Sparta as slight favourites. Chlapík's early goal (0:32) set the tone exactly as our analysis suggested. Sparta's shot conversion rate of 15.4% far exceeded their season average, while Plzeň's 6.3% was uncharacteristically poor. The series moves to a decisive Game 7 in Plzeň.

More archived articles will appear here as the playoffs progress.