A raw sharpe-ratio ignores how many strategies you tried to get it. The deflated version asks: given that I ran 200 trials on 5 years of data with returns this skewed and fat-tailed, how impressive is a Sharpe of 1.4 really?
Two effects dominate. Trials: the expected maximum Sharpe from N zero-edge trials grows roughly with the square root of the logarithm of N, so 200 trials on 5 years can produce an apparent 0.9 from nothing. Non-normality: negative skew and high kurtosis widen the sampling error, so the same Sharpe carries less evidence.
Used honestly it rejects most research. That is the point. If you cannot state your trial count, you cannot compute it, which is itself a finding about your process.
Related: probabilistic-sharpe-ratio, sharpe-ratio, multiple-testing, minimum-backtest-length