State the hypothesis before seeing results
We write each idea as a falsifiable hypothesis with a pass threshold set in advance. We log every attempt, failures included, so we know how many tries a good result cost.
- Research log
- Pre-set thresholds
Quantitative trading research on the Vietnamese market
VietHedge Terminal
Every trading idea, however sensible it sounds, is only a hypothesis until it passes statistical testing. We test on data that was not used to build the strategy.
Most ideas do not get past this step.
Research
How we research
Markets are noisy. A good-looking backtest can be luck, or the result of trying too many settings. These four principles help us tell a real edge from luck.
Compare against chance
We compare each strategy with hundreds of versions that enter at random, with the same frequency and holding time. A result counts only if very few random versions reach it.
- Permutation test
- Bootstrap
Adjust for the number of tries
The more settings you try, the easier it is to find a good one by luck. So we adjust the Sharpe ratio for the number of tries, and we estimate the probability that a strategy merely overfits past data.
- Deflated Sharpe ratio
- Probability of backtest overfitting (PBO)
Use only data available at decision time
Every decision in a simulation uses only the data available at that moment. We check this by truncating the data at each day and recomputing. We choose parameters on one period and judge the strategy on the next, which the model has never seen.
- Point-in-time data
- Walk-forward
- Out-of-sample
Process
From idea to real capital
Hypothesis
An idea with an economic or behavioural reason, written as something testable.
Backtest
A simulation on data from 2020, with fees, tax and a liquidity cap.
Statistical tests
Against chance, adjusted for the number of tries, checked out of sample.
Paper trading
Runs on live market prices every day, without capital.
Real capital
Only a small share of ideas get this far.
We share real trading results only when working directly with partners.
Results
Strategies that passed testing
The figures below are backtest results over the period shown for each strategy, not real trading results. Each strategy's page has the full figures and method.
Assumptions
Simulation assumptions
We use minute prices for VN30 futures and for the ETF, and daily prices for the bank stocks. The data starts in 2020.
An order fills only at a price after the signal. Each day we recompute the weights between the component strategies using data up to the end of that day. This avoids look-ahead bias.
For stocks and the ETF we charge a fee of 0.15% per side and a sale tax of 0.1%. For derivatives we apply the fees and tax under the current rules.
We chose the futures short book of the bank strategy on data before 2024, so the period from 2024 onward is out-of-sample.
No order exceeds 10% of market volume. From that limit we estimate the largest capital each strategy can run, for example about VND 40 billion for the derivatives strategy.
We simulate the bank strategy at closing prices and do not model slippage, so results from real trading may be lower.
We discuss signal details when working directly with partners.
Work with us
We work with funds, securities companies and individual investors.
Research partnership
We supply data, run backtests and build strategies on request.
Investment partnership
We run a strategy on a partner's capital.
VietHedge Terminal is a data platform for education and research in quantitative trading.
Not investment advice. Past performance does not guarantee future results.