Signatures provide a convenient and widely accepted method of authentication, however they are prone to attack by forgery. This can be mitigated to an extent by analysing both the static and dynamic biometric aspects of construction, however the possibly for accurate forgery from a static image of a genuine signature still exists. In this study we explore initial forgery accuracy of a range of genuine signatures and how accuracy changes as a forger receives feedback from a commercial signature engine in terms of a 'match score'. We also explore the effects of genuine signature complexity on forgery performance alongside the image size of the genuine signature to be forged. Our results show that forgers are able improve performance over time on simple signatures (including those with less pen travel distance) and that a magnified genuine sample enables more accurate forgery for these class of signatures. More complex signatures result in lower forged verification scores and irregular patterns of improvement across the five forgeries. Overall verification match scores were typically less than 80% for most attempts regardless of signature complexity, thus indicating the resilience of dynamic systems to unskilled forgery attempts.