> Probability of Haaland scoring goals (inc. penalties) as a normal distribution + implied points including bps, discrete buckets
For **Haaland vs Coventry specifically**, the cleanest market-implied model is actually a **discretized normal**, calibrated to today’s scoring markets.
The market has Haaland at about **1.43 anytime scorer** (~69.9% raw implied probability) and **6.50 for a hat-trick** (~15.4% raw implied probability). ([Football Whispers][1]) Those two points imply a fairly fat-tailed scoring distribution, so I fit a latent normal that reproduces both approximately:
**Goals ~ N(μ ≈ 1.18, σ ≈ 1.30)**, then bucketed into integer goal counts.
### Haaland goal distribution vs Coventry
| Goals | Probability | Raw FPL pts* | Estimated pts incl. bonus |
| -----: | ----------: | -----------: | ------------------------: |
| **0** | **30.1%** | 2 | **~2.0** |
| **1** | **29.8%** | 6 | **~7.2** |
| **2** | **24.8%** | 10 | **~12.6** |
| **3** | **11.7%** | 14 | **~17.0** |
| **4** | **3.1%** | 18 | **~21.0** |
| **5** | **0.5%** | 22 | **~25.0** |
| **6+** | **<0.1%** | 26+ | **~29+** |
*Assumes 60+ minutes, no assist/card/penalty miss. Forwards get **4 points per goal**, and FPL awards 1–3 bonus to the top three BPS performers. ([Premier League][2])
My bonus assumptions are roughly **+1.2 after one goal, +2.6 after two, +3 after 3+**. They’re estimates, because bonus depends on what everyone else does; a penalty goal also has a weaker BPS impact than an open-play goal under the current system.
The key cumulative probabilities are therefore:
* **Blank (0 goals): ~30%**
* **Exactly one goal: ~30%**
* **2+ goals: ~40%**
* **3+ goals: ~15%**
That last number lines up intentionally with the quoted **6.50 hat-trick market**. The anytime market also lines up at about **70% to score at least once**. ([Football Whispers][1])
### What this means for your situation
The modal outcomes are actually **0, 1 or 2 goals**, not “Haaland automatically braces”.
Using those probabilities and the bonus assumptions, his estimated mean is about:
**~8.7 FPL points**
which means roughly:
* **Haaland (C): ~17.3 expected**
* **Haaland TC: ~26.0 expected**
But the distribution is much more useful than the average:
| Haaland outcome | Chance | Captain score | TC score |
| ------------------- | -----: | ------------: | -------: |
| **Blank ~2 pts** | 30% | **4** | **6** |
| **1 goal ~7 pts** | 30% | **~14** | **~22** |
| **2 goals ~13 pts** | 25% | **~25** | **~38** |
| **3+ goals ~17+** | 15% | **~34+** | **~51+** |
So for us, the battlefield is pretty clear.
**~60% of the distribution is Haaland scoring zero or one goal.** Those outcomes are very survivable with **Bruno (C)** plus Semenyo/Cherki/MGW/Arsenal returns.
The dangerous ~40% is **2+ goals**, and the genuinely catastrophic tail against the TC teams is the **~15% hat-trick-or-better region**.
Also, this incorporates penalties as part of the goal count — i.e. a converted Haaland penalty is simply another goal. It does **not** price in the negative points from a missed penalty.
[1]:
https://footballwhispers.com/blog/manchester-city-vs-coventry-city-backing-haaland-to-run-riot-against-newly-promoted-premier-league-club/ "Man City vs Coventry: Tempting Erling Haaland odds for Premier League clash"
[2]:
https://www.premierleague.com/en/news/2174909/fpl-basics-explained-scoring-points?utm_source=chatgpt.com "FPL basics explained: Scoring points"
> Let's run a sample variance given all Manchester City games from the last 2 seasons (intentionally leaving the 23/24 outlier out) excluding games missed
Yes — excluding games he **did not play at all** is the right way to do it.
For a clean FPL-relevant sample, I used Haaland’s **Premier League appearances in 2024/25 and 2025/26 only**, leaving 2023/24 out exactly as you requested. He made **31 PL appearances in 2024/25** and **35 in 2025/26**, so the sample is **66 matches played**. He scored 22 and 27 goals respectively. ([StatMuse][1])
Across those 66 appearances, his goal-count distribution was:
| Goals in match | Matches | Empirical probability |
| -------------: | ------: | --------------------: |
| 0 | 30 | **45.5%** |
| 1 | 25 | **37.9%** |
| 2 | 9 | **13.6%** |
| 3 | 2 | **3.0%** |
| **2+** | **11** | **16.7%** |
The underlying match logs support that distribution: 2024/25 contained 15 blanks, 12 one-goal games, two braces and two hat-tricks; 2025/26 contained 15 blanks, 13 one-goal games and seven braces. ([Statbunker football][2])
So the historical sample gives:
**Mean goals/appearance = 49 / 66 = 0.742**
and the **sample variance** is:
**s² ≈ 0.656**
with sample standard deviation:
**s ≈ 0.810 goals**
That is quite revealing for the Coventry concern.
The market-style model we were using gave roughly **40% for 2+ goals today**. But over the last two completed PL seasons, Haaland actually scored **2+ in only 16.7% of the matches he played**.
So a 40% brace-or-better probability would be about:
**2.4× his recent empirical base rate.**
That doesn't mean 40% is impossible — Coventry at home is obviously much easier than his average opponent over a 66-match sample — but it shows just how aggressively the market/model is upgrading him for this particular fixture.
There is another interesting feature: the sample is **under-dispersed relative to a Poisson**. With mean λ = 0.742, a Poisson would have variance 0.742; observed sample variance is only **0.656**. In other words, over these two seasons his scoring has actually been slightly **less volatile than a basic Poisson goal model would expect**.
Using his empirical two-season distribution directly:
**P(blank) ≈ 45.5%**
**P(exactly 1) ≈ 37.9%**
**P(2+) ≈ 16.7%**
So roughly **83.3% of his appearances ended with zero or one goal**.
[1]:
https://www.statmuse.com/fc/ask?q=erling+haalands+24%2F25+stats&utm_source=chatgpt.com "Erling Haalands 24/25 Stats | StatMuse"
[2]:
https://statbunker.com/players/SeasonMatches?comps_id=-1&comps_type=EPL&player_id=60023 "Erling Haaland Player matches Premier League"
As you can see, I did my homework. I think my team will outscore Haaland (C), TC is a different story. I will wildcard when I have an idea which players are actually good, instead of randomly picking up players like Gross and Wissa who may or may not return.
I did not want to sell Mbeumo for Cherki, so I got rid of Szobo instead. So far, that turned out to be good.
