I've done calorie counting a while back. It does not have to be very accurate. Getting more accurate is easier and more efficient after you do it for a while.

I've done calorie counting a while back. It does not have to be very accurate. Getting more accurate is easier and more efficient after you do it for a while. I realized that what I eat is fairly stable. There may be 20 different things I eat regularly. Eventually, I started memorizing the most often used ones. I noticed that there is a pattern to which types of foods have what amounts of calories. For example, fresh fruit and many non-starchy veggies tend to be at around 50kcal/100gr. Starchy veggies got for about 100kcal/100gr. Meats are at around 250kcal/100gr. And so on. Using this, I was able to get estimates for things I don't eat often without needing to Google everything. I did look up things I ate a few times in order to test my estimates. It is not super-accurate. However, writing it down is more important that writing it down accurately to the last calories. Many parts of this process were not accurate. I weighed things roughly using an analog scale. I often estimated portion size by eye, as I got a sense of how much 200grams of meat is. I counted the calories using estimates a lot of time. It worked nonetheless. After 6 months, my weight loss vs. estimated weight loss from calories spreadsheet was within 1kg! I was blown away by how close this was. It makes sense since things like your base metabolic rate and actual calories expenditure also varies, so super-accurate record keeping of food-in was not necessary. You're right that being off by 100 calories systematically will force a bias. The trick is to be off in one direction one day, and in the other the other day. This way, the error mostly cancels out.