Fix a planning period \(T>0\) , an arbitrage free pricing model providing us with the current numerical price of a general random pay-off with maturity at T, and a coherent risk measure. Denote by \(\mathbb {Q}\) and \(\mathbb {P}\) the risk neutral and the physical probability measure, respectively. Denote finally by \(d\mathbb {Q}/d\mathbb {P}\) the Radon-Nikodym derivative of \(\mathbb {Q}\) with respect to \(\mathbb {P}\) , and by \( I_{0}\ge 0\) the essential infimum of \(d\mathbb {Q}/d\mathbb {P}\) . Consider \(\varepsilon >0\) and \(k>\varepsilon \) . A marketed claim \(y\ge 0\) with a current price lower than \( \varepsilon \) , an expected pay-off higher than k, and a coherent risk lower than \(-k\) is said to be a \(\left( k,\varepsilon \right) -strongly\_golden\_strategy\) ( \(\left( k,\varepsilon \right) -SGS\) ). Since this investment strategy does not entail any liabilities ( \(y\ge 0\) ), the potential capital losses are bounded by \(\varepsilon \) , whereas the potential capital earnings \( y-\varepsilon \) are significant if \(k/\varepsilon \) is large enough. Most risk measures do not allow for the existence of any \( \left( k,\varepsilon \right) -SGS\) if \(k/\varepsilon \) is large, but the expectile does if \(I_{0}\) is close to zero. In particular, if \(I_{0}=0\) , then there exists a \(\left( k,\varepsilon \right) -SGS\) for every \(0<\varepsilon <k\) , and the equality \( I_{0}=0\) frequently holds in continuous time (Black-Scholes-Merton model, Heston model, etc.). This might be a significant difference between the expectile and other coherent risk measures such as the conditional value at risk. In addition, it is proved that the strategy y above is easy to detect in practice and may be often composed of simple derivative securities. Thus, though this paper is only theoretical, providing simple criteria for the practical detection of a \(\left( k,\varepsilon \right) -SGS \) facilitates the implementation of future empirical tests.
