The momentum puzzle, i.e., stocks that have performed better in the past tend to perform better in the future, has been a constant challenge to classic finance theory. Prior research has failed to provide valid risk-based explanations because winner portfolios do not exhibit higher risk characteristics. Without a convincing risk explanation, the persistence of momentum profit is a violation of the efficient market hypothesis. Today, the momentum puzzle remains one of the very few major anomalies that cannot be explained by Fama-French factor models. I find prior empirical efforts to measure momentum profits and its sources are contaminated by the state of the market during both formation and holding periods. By looking into different market states, classified by both traditional and non-traditional bull and bear market definition, I find the key to at least partially solve the momentum mystery. Momentum stocks are riskier when formed in bull market, and momentum profit is much higher in continuation of market than reverses of market condition, lending empirical support to a risk-based explanation. My definition of market states is essentially based on the risk premium of major risk factors. When market risk is considered a risk factor, if realized market risk premium is positive, it is a bull market; when size is considered a proxy for risk factor, if SMB (small minus big risk premium) is positive, it is a bull market; when valuation (book-to-market) ratio is a proxy for risk factor, if HML (High-minus-Low risk premium) is positive, it is a bull market. This paper also explores simulations using models based on the positive relationship between risk and return. The simulation result confirms that at least part of the momentum profit can be explained by risk.