How Often Do HKJC Favourites Actually Win? 5 Seasons of Data

Everyone has an opinion about favourites. Some punters only bet them, others refuse to. But what do the numbers actually say? Renavon's hkjc_race_results dataset has every race result since 1977, with win odds for every runner. That's enough data to settle this definitively.

I looked at the last 5 complete racing seasons (September 2021 to April 2026): 3,955 races and 360,000+ individual race entries.

The headline numbers¶

The market favourite — the horse with the lowest win odds at the jump — wins 31.2% of races. That's roughly one in three.

If you include places, favourites finish in the top three 61.7% of the time. The market is clearly not random. But is it profitable?

The median favourite goes off at odds of 3.0 (i.e. $3 for a $1 stake). If you flat-bet $1 on every favourite across these 3,955 races, you'd be down $606 — a return on investment of roughly -15%. The HKJC's takeout wins.

Win rate by odds bracket¶

Not all favourites are equal. A 1.3 odds-on certainty is a very different proposition from a 6.0 favourite in a wide-open handicap.

WITH favourites AS (
    SELECT
        race_date, race_number, horse_name, odds,
        finishing_position, is_winner, is_top_three,
        ROW_NUMBER() OVER (
            PARTITION BY race_date, race_number
            ORDER BY odds ASC
        ) AS odds_rank
    FROM renavon_hkjc_race_results.main.hkjc_race_results
    WHERE race_date >= '2021-09-01'
      AND finishing_position > 0
      AND odds > 0
)
SELECT
    CASE
        WHEN odds < 1.5 THEN '< 1.5 (odds-on)'
        WHEN odds < 2.0 THEN '1.5 - 1.9'
        WHEN odds < 3.0 THEN '2.0 - 2.9'
        WHEN odds < 4.0 THEN '3.0 - 3.9'
        WHEN odds < 5.0 THEN '4.0 - 4.9'
        WHEN odds < 7.0 THEN '5.0 - 6.9'
        ELSE '7.0+'
    END AS odds_bracket,
    count() AS races,
    sum(CASE WHEN is_winner THEN 1 ELSE 0 END) AS wins,
    round(sum(CASE WHEN is_winner THEN 1 ELSE 0 END) * 100.0 / count(), 1) AS win_pct,
    round(sum(CASE WHEN is_winner THEN odds - 1 ELSE -1 END) * 100.0 / count(), 1) AS roi_pct
FROM favourites
WHERE odds_rank = 1
GROUP BY odds_bracket
ORDER BY min(odds);

A few things jump out:

• Odds-on favourites (< 1.5) win 75% of the time. Three in four. But the ROI is still -5.3% because when they lose, you lose your entire stake while winning pays very little.
• The sweet spot might be 5.0 - 6.9. This is the only bracket that shows a positive flat-bet ROI (+6.2%), though the sample is small (133 races). These are "vulnerable" favourites where the market has less confidence — and apparently misprices them slightly.
• The most common favourite bracket is 2.0 - 2.9 (1,442 races). These win about a third of the time, which feels right intuitively — and the -11.1% ROI is roughly in line with the HKJC's win pool takeout.

Does venue matter?¶

Sha Tin and Happy Valley are very different tracks. Does that affect how often favourites get home?

A small but consistent difference: favourites win about 2.7 percentage points more often at Sha Tin. Happy Valley's tighter track and shorter straights create more upsets — barriers and racing luck matter more.

Season-by-season consistency¶

Is the favourite win rate stable, or does it swing wildly?

WITH favourites AS (
    SELECT
        race_date, race_number, odds, is_winner,
        CASE
            WHEN race_date >= '2025-09-01' THEN '2025/26'
            WHEN race_date >= '2024-09-01' THEN '2024/25'
            WHEN race_date >= '2023-09-01' THEN '2023/24'
            WHEN race_date >= '2022-09-01' THEN '2022/23'
            WHEN race_date >= '2021-09-01' THEN '2021/22'
        END AS season,
        ROW_NUMBER() OVER (
            PARTITION BY race_date, race_number
            ORDER BY odds ASC
        ) AS odds_rank
    FROM renavon_hkjc_race_results.main.hkjc_race_results
    WHERE race_date >= '2021-09-01'
      AND finishing_position > 0
      AND odds > 0
)
SELECT
    season,
    count() AS races,
    sum(CASE WHEN is_winner THEN 1 ELSE 0 END) AS fav_wins,
    round(sum(CASE WHEN is_winner THEN 1 ELSE 0 END) * 100.0 / count(), 1) AS win_pct,
    round(sum(CASE WHEN is_winner THEN odds - 1 ELSE -1 END) * 100.0 / count(), 1) AS roi_pct
FROM favourites
WHERE odds_rank = 1
GROUP BY season
ORDER BY season;

Remarkably stable. The win rate bounces between 29% and 33% — a narrow band. The ROI varies more (-10.9% to -19.8%), driven by whether the winners come from the shorter or longer odds end of the spectrum.

Race class matters more than you'd think¶

One of the most interesting splits. Do favourites win more often in higher-class races?

The pattern is clear: the higher the class, the more often the favourite wins. Group 1 favourites win 72% of the time — the cream really does rise in the top races. Class 5 handicaps, where all the horses are of similar (low) ability, produce the most upsets.

This makes intuitive sense. In a Group 1, the field includes one or two genuinely superior horses. In a Class 5 handicap, the weights are designed to equalise the field and any of 14 horses might win.

The bottom line¶

If you're betting on Hong Kong racing:

1. Favourites win about 31% of the time. Reliable, but not enough to profit on flat bets.
2. Odds-on shots (< 1.5) win 75% of the time but the ROI is still negative. You're paying a heavy premium for reliability.
3. Sha Tin favours favourites more than Happy Valley. The wider track rewards the better horse.
4. Group races are the most predictable. If the favourite is a genuine class above, back it with confidence.
5. Flat-betting favourites will always lose money due to the HKJC's takeout. You need an edge, not a system.

Run the analysis yourself¶

All the SQL queries above run directly on Renavon's hkjc_race_results dataset via DuckDB/MotherDuck. The dataset covers every race since 1977 — 360,000+ entries — and is updated after every race meeting.

Want to dig deeper? Try filtering by distance, surface type, or specific jockeys and trainers. The data is there.