R resample results are not meaningful at (assess =1, and assess = 6)

[Shafi_q]

I am using the sliding_window() or rolling_origin() from resample R package . For the one-step-ahead forecast, I used assess = 1, and six steps ahead forecast assess = 6. First, I tune the models then passed these tuned models to fit_resamples() . Get the RMSE from these slices and plot against the time. However, the issue is that I get a larger RMSE at assess= 1 than assess = 6 which is counter-intuitive obviously it should be a larger RMSE at assess =6. I have given below reproducible codes with hypothetical data and graphs. I do not know what is wrong. Is there is an issue with my method? Please help.

library(Quandl)
library(tidymodels)
library(resample)
library(tidyverse)
library(lubridate)
library(timetk)
df1 <- Quandl(code = "FRED/PINCOME",
              type = "raw",
              collapse = "monthly",
              order = "asc",
              end_date="2017-12-31")
df2 <- Quandl(code = "FRED/GDP",
              type = "raw",
              collapse = "monthly",
              order = "asc",
              end_date="2017-12-31")

per <- df1 %>% rename(PI = Value)%>% select(-Date)
gdp <- df2 %>% rename(GDP = Value) 

data <- cbind(gdp,per)

data1 <- tk_augment_differences(
  .data = data,
  .value = GDP:PI,
  .lags = 1,
  .differences = 1,
  .log = TRUE,
  .names = "auto") %>%
  select(-GDP,-PI) %>%
  
  rename(GDP = GDP_lag1_diff1,PI = PI_lag1_diff1) %>% 
  drop_na()


lag_period <- c(1:5)

data_pre_full <- data1   %>%
  # add lags
  tk_augment_lags(
    .value =  GDP:PI, 
    .lags = lag_period)  


data_prepared_tbl <- data_pre_full %>%
  dplyr :: select( -PI )  %>%  
  
  drop_na() 

splits <- timetk::time_series_split(data_prepared_tbl, assess = 70, cumulative = TRUE)

recipe_spec <- recipe(GDP ~ ., data = training(splits)) #%>%



# *Resampling----- 
resamples_tscv <- sliding_window(data_prepared_tbl, 
                                 lookback = 100, 
                                 assess_stop = 1, step = 2
)


plan_org <- resamples_tscv %>% tk_time_series_cv_plan()

plan_tr1  <- plan_org %>% 
  filter(plan_org$.key == "testing")
plan <- plan_tr1 %>% select(Date,.key,.id,GDP)

# * SVM ----

wflw_fit_svm <- workflow() %>%
  add_model(
    spec = svm_rbf(mode = "regression") %>% set_engine("kernlab")
  ) %>%
  add_recipe(recipe_spec %>% update_role(Date, new_role = "indicator")) %>%
  fit(training(splits))

# Fit resamples
keep_pred <- control_resamples(save_pred = TRUE, verbose = TRUE)
set.seed(20201024)
# Fit resamples
rf_res <- fit_resamples(wflw_fit_svm,        
                         resamples = resamples_tscv, control = keep_pred)
rf_res

pr <- rf_res %>% 
  # grab specific columns and resamples
  pluck(".predictions")

pred1<- rf_res %>% 
  # combine ALL predictions
  collect_predictions()

pred_final1 <- cbind(pred1,plan)
write.csv(pred_final1, file = "pred.SVM1.csv")
# Summarize all metrics
rf_res %>% 
  collect_metrics(summarize = TRUE)

rmse <- rf_res %>%  collect_metrics(summarize = FALSE)
rm  <- rmse %>% 
  filter(rmse$.metric == "rmse")
write.csv(rm, file = "rmse_1.csv")