(Listen to this article)


The allocation of research funding represents one of the most consequential decisions in scientific policy, specially these days with some of the most drastic cuts happening in available funding. One major approach by reviewers and groups in charge of fund apprpriation is to often prioritize researchers/research groups with established track records, operating under the assumption that past success predicts future breakthroughs. However, statistical principles suggest that maximizing the diversity and volume of funded research—rather than concentrating resources among previously successful investigators—may yield superior outcomes for scientific discovery.


The Numbers Game

One could argue that the relationship between research funding and breakthrough discoveries follows a counterintuitive statistical pattern. When we observe exceptional research outcomes in any given year, we naturally attribute these successes to the superior abilities of individual researchers. However, this attribution obscures a more fundamental truth: the number of outstanding discoveries depends primarily on the total number of research projects initiated, not on our ability to identify the most talented investigators.

Consider this from a probability perspective. Even assuming some researchers genuinely possess superior capabilities, breakthrough discoveries remain subject to substantial random variation. Environmental factors, technological timing, collaborative serendipity, and countless unforeseen variables contribute to research success. In such a system, the most reliable method for increasing exceptional outcomes is increasing the total number of funded projects—expanding the denominator rather than trying to optimize the numerator.

The survivorship bias that leads us to celebrate successful researchers while ignoring those who produce incremental results depends on the size of the initial population. If we fund 100 researchers, we might observe 5 breakthrough discoveries. If we fund 1,000 researchers, we’re likely to see closer to 50 breakthroughs. This is not because we’ve become better at identifying talent, but because we’ve increased the statistical opportunities for exceptional outcomes to emerge.

So concentrating funding among previously successful researchers, despite its intuitive appeal, may actually reduce the total number of scientific breakthroughs. We’re optimizing for past performance in a system where future success is largely stochastic.


Regression to the Mean - Today’s Stars Become Tomorrow’s Average

The statistical concept of regression to the mean provides a second compelling argument against concentrating funding among elite researchers. In any system with random variation, extreme outcomes in one period tend to move toward the average in subsequent periods. Applied to research productivity, this means that investigators who achieve breakthrough discoveries are statistically expected to produce more typical results in future research cycles.

This regression occurs not because successful researchers lose their abilities, but because extraordinary outcomes represent statistical outliers that are inherently difficult to replicate. The researcher who made a revolutionary discovery benefited from a unique constellation of circumstances (favorable timing, unexpected collaborations, technological convergences) that cannot be deliberately recreated. Meanwhile, the broader population of researchers continues to represent a pool of potential breakthroughs waiting for their own favorable alignment of conditions.

Time, in a way, eliminates the effects of randomness. The researcher who appeared exceptional based on a single breakthrough will, over extended periods, likely demonstrate performance closer to the population average. Funding agencies that concentrate resources among previously successful researchers may therefore be systematically investing in regression rather than continued excellence.


Practical Implications

These statistical principles suggest that current funding concentration trends work against scientific progress. Rather than trying to identify the researchers most likely to achieve breakthroughs, an essentially impossible task given the role of randomness, funding agencies should maximize the number of research opportunities. This approach treats scientific discovery as a statistical sampling problem where the goal is coverage rather than precision.

The optimal strategy becomes clear: fund as many qualified researchers as possible, regardless of their track records. The researcher who has never made a major discovery may be statistically more likely to produce the next breakthrough than the investigator who achieved recent success. In the unpredictable landscape of scientific discovery, quantity generates its own quality through the fundamental workings of probability.

Note: This perspective doesn’t diminish the value of scientific talent or suggest that all researchers are equivalent. It just acknowledges that in systems dominated by randomness, our ability to predict individual success is limited, while our ability to predict population-level outcomes is more robust and reliable. We cannot identify which specific researcher will make the next breakthrough, but we can be confident that more funded researchers will produce more breakthroughs, as simple as that.