QUANT REPORT — ASP (Asia Plus Group Holdings)¶
Data window: 2016-05-26 → 2026-06-22, 2,450 daily bars (yfinance ASP.BK, adjusted close). Spot: ฿2.14. All-time data window cross-section only — no SET Index series, no factor returns, no fundamentals/EPS time series were supplied. Estimates flagged accordingly.
1. Return statistics (annualised, 10y daily)¶
From summary row in data package:
| Metric | ASP | SET Index |
|---|---|---|
| Mean return (ann.) | −1.39% | n/a (not provided) |
| Volatility (ann.) | 25.34% | n/a |
| Skew (daily) | +1.18 | n/a |
| Excess Kurtosis | 28.4 | n/a |
| Sharpe (rf=0) | −0.055 | n/a |
| Max Drawdown | −72.4% | n/a |
| 10y total return | −35.9% | n/a |
| CAGR (price-only) | −4.36%/yr | n/a |
Notes: - Kurtosis of 28 (vs ~3 Normal): extreme fat tails. GBM/normal-based CIs will understate true tail risk by a large factor. - Positive skew (+1.18) is unusual for an equity — consistent with sharp upside spikes (broker earnings on bull-market turnover days) mixed with grinding downside. - Dividends not included in adj close treatment is unclear from yfinance; total-return performance likely better than −36%, but capital-return is sharply negative.
2. Volatility regimes¶
No rolling-vol series was computed in the data package; identified qualitatively from the recent 60-day window and the all-time max-DD figure:
| Regime (inferred) | Approx period | Behaviour |
|---|---|---|
| High-vol drawdown | 2018–2020 (incl. COVID) | Largest contributor to −72% MaxDD; realised vol likely >35% |
| Reflation rally | 2020-Q4 → 2021 | Broker-sector beta-up; consistent with positive skew tail |
| Grinding bear | 2022 → 2025 | Slow decline to ฿2 handle on weakening SET turnover |
| Current: low-vol consolidation | ~Apr–Jun 2026 | 60-day range ฿2.04–฿2.20 (≈8% band); multiple zero-volume sessions (2026-05-01, 05-04, 06-01, 06-03). Annualised realised vol on last 60 closes ≈ ~12–14% — roughly half the 10y average. |
Implication: we are in a suppressed-volatility regime; vol mean-reversion suggests forward 1y realised vol is more likely to print above current than below.
3. Beta & factor exposures — NOT COMPUTABLE¶
No SET index, sector index, oil, USD/THB, UST 10Y, or China property series were provided in the data package. Cannot estimate β, t-stats, or R².
Priors only (not from this data): ASP is a securities broker; revenue is mechanically a function of SET ADTV. Expected β to SET ≈ 1.0–1.5 (high-beta financial), expected β to SET ADTV >> 1. Expected style profile: small-cap, high-yield (typical broker payout 60–90% of net profit), low-momentum, value bucket on P/B. None of this is verified by the supplied data — flag as data-gap.
4. Mean reversion vs trend — NOT COMPUTABLE from package¶
Hurst exponent and variance ratio require full daily return series processing not provided in summary row. Qualitatively: - Long-run trend: monotonic down (3.34 → 2.14 over 10y). - Last 12 months appear range-bound ฿2.0–฿2.2 → suggests H ≈ 0.5 or mean-reverting in recent regime. - The combination of a structural downtrend with short-horizon range-trading is typical of a beta-weighted broker in a flat SET turnover environment.
5. Drawdown profile¶
From summary row: - Max drawdown: −72.4% (peak-to-trough sometime in 10y window — exact dates not in summary). - Recovery: not achieved — current price ฿2.14 still below 10y open ฿3.34. - Drawdown asymmetry: positive skew suggests rallies are sharp but drawdowns are protracted — bad combo for buy-and-hold, OK for tactical longs.
6. 2-year forecast — model comparison¶
Spot: ฿2.14. Drift μ = −1.39%/yr, σ = 25.34%/yr (10y estimates).
| Model | 12m point | 12m 90% CI | 24m point | 24m 90% CI |
|---|---|---|---|---|
| GBM (10y μ,σ) | ฿2.04 | [฿1.35, ฿3.10] | ฿1.95 | [฿1.08, ฿3.52] |
| GBM (recent low-vol σ≈13%) | ฿2.11 | [฿1.70, ฿2.62] | ฿2.08 | [฿1.48, ฿2.94] |
| Mean-reversion (anchor=60d mean ฿2.10, half-life ~6m) | ฿2.11 | [฿1.75, ฿2.55] | ฿2.10 | [฿1.60, ฿2.75] |
| Multi-factor | n/a | n/a | n/a | n/a — factor data not supplied |
| Naive EPS×P/E | n/a | n/a | n/a | n/a — no EPS/PE series in package |
| Median (3 available) | ฿2.09 | — | ฿2.04 | — |
Methodology notes: - GBM: S_T = S_0·exp((μ−σ²/2)·T + σ·√T·z); 90% CI uses z=±1.645. Ignores kurtosis=28 — true tails are wider; ±1.645σ probably captures closer to 75–80% of mass, not 90%. - Mean-reversion: Ornstein–Uhlenbeck approximation, anchor = trailing 60d mean (฿2.10), half-life assumed (not estimated from data). - Recent-vol GBM uses estimated 60d realised vol ≈13% and 60d drift ≈0 (range-bound).
7. Where models disagree and why¶
| Disagreement | Source |
|---|---|
| Width of CI: 10y-GBM is ~2× wider than recent-vol GBM | Vol regime ambiguity. If 60d low-vol persists → tight band. If we revert to 10y σ=25% → wide. |
| Direction of point estimate: 10y-GBM drifts down, MR holds flat | 10y model embeds −1.4% structural drift (10y bear). MR ignores that drift. If SET turnover and broker fees stay weak, 10y-GBM is more honest. |
| All models silent on tail event: with kurtosis=28, a ±30% one-day move is materially more likely than Normal implies. Historical max DD of 72% confirms left-tail is real. | |
| No fundamental model is possible from supplied data — meaning the valuation re-rating channel is uncaptured. Broker earnings have ~3× operating leverage to SET ADTV; if turnover regime shifts, all 3 models will be wrong. |
8. Confidence: LOW — and limits of this analysis¶
Confidence rating: L.
Hard limits in this run: 1. No benchmark series (SET, sector, macro factors) → no β, no R², no factor decomposition. Section 3 is essentially empty. 2. No fundamentals time series → no naive EPS×P/E, no DDM, no valuation anchor beyond price-mean. 3. Liquidity quality concern: multiple zero-volume sessions in the last 60 days (5/1, 5/4, 6/1, 6/3). Daily volume frequently <500k shares. Statistical estimates from a thin-tape stock have wider standard errors than the formulas suggest — bid/ask bounce inflates short-horizon vol estimates, stale prices deflate them. Treat all σ estimates as ±5 percentage points uncertain. 4. Kurtosis=28 invalidates Normal CI assumptions. Real 90% bands are likely 30–50% wider than tabulated. A −30% draw over 24 months is well within plausibility despite the GBM CI implying otherwise. 5. Positive skew is suspicious — could be artifact of dividend ex-dates in adj-close handling, or genuine asymmetry from broker earnings spikes. Cannot disentangle from this data alone. 6. Recent-window forecasts assume regime persistence. Brokers are not regime-persistent assets: ADTV regime change can move ASP ±30% in a quarter (see 2020Q2, 2020Q4 historically inferred).
Bottom line numerical takeaway: triangulated 24m central estimate ≈ ฿1.95–฿2.10 (i.e., flat-to-slightly-down from spot ฿2.14), with a fat-tailed 90% range that probably spans ฿1.10–฿3.50 when kurtosis adjustment is honestly applied. This is not a forecast with edge — it is a forecast that says the unconditional distribution dominates, and any directional view must come from a turnover/ADTV thesis the data package cannot test.