"Learning with tens of billions of ticks" is padding. The correct way to count learning progress

In AI product descriptions, you may see expressions like “trained on billions of tick data.” It’s overwhelming to hear billions, but please convert it calmly.

A tick is one update of the rate, and with actively traded currency pairs, tens to hundreds of ticks occur per minute. In other words, billions of ticks equal only several years’ worth of price movement when converted to years. Writing the same data as “billions of ticks” or “hundreds of millions of candlesticks” gives the impression of a four-digit difference even though the content is the same. By using large units, the amount of learning can be easily inflated.

Furthermore, more learning data isn’t always better. Having the rules of a market from ten years ago learned in large quantities can add noise to the current market. The question isn’t “how many items” but “from which period and at what granularity of data.”

The learning amount for ‘GENSEN-AI’ is up to 100,000 candlesticks. It won’t be presented in terms of ticks to appear larger. At H1, 100,000 candlesticks correspond to several years, and we state the granularity of learning (candlesticks) and the number as-is. How the learning amount is stated serves as a benchmark when comparing products. If the unit is ticks, convert it back to candlesticks for a proper comparison. That alone restores the numbers to their correct scale.